[Bug c++/19748] New: aggressive no-inline options still cause inlining

[Bug c++/19748] New: aggressive no-inline options still cause inlining

[yuri at tsoft dot com]

I am trying to compile with absolutely no inlining. And still get some functions
inlined. Though somewhat less than only -no-inline.

My compile/link options include:
-fno-implement-inlines -fno-default-inline -finline-limit=0 -fno-default-inline
-fno-inline
and also -O3 -no-guess-branch-probabilities

If I want to profile specific functions for data cache misses no-inline can be
very useful.

Yuri

-- 
           Summary: aggressive no-inline options still cause inlining
           Product: gcc
           Version: 3.3.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: yuri at tsoft dot com
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748

[Bug c++/19748] aggressive no-inline options still cause inlining

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-02-02 01:56 -------
-O3 enables -finline-functions, you want -fno-functions-inlines but -fno-inline should not enable 
inlining period (except for alwaysinline functions).

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748

[Bug c++/19748] aggressive no-inline options still cause inlining

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-02-02 01:57 -------
Also note sometimes when a function is pure/const it can be removed which is why it might act as 
inlining.

Do you have a simple example?

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748

[Bug c++/19748] aggressive no-inline options still cause inlining

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-02-02 02:00 -------
(In reply to comment #2)
> Also note sometimes when a function is pure/const it can be removed which is why it might act as 
> inlining.
This is an example where we remove the function call so it looks like inlining:
void f(void) {}
void g(void)
{
  f();
}
But we are not inlining here at all, we just removed the function as the function is pure/const.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748

[Bug c++/19748] aggressive no-inline options still cause inlining

[yuri at tsoft dot com]

------- Additional Comments From yuri at tsoft dot com  2005-02-02 02:07 -------
(In reply to comment #2)
> Also note sometimes when a function is pure/const it can be removed which is
why it might act as 
> inlining.
> 
> Do you have a simple example?
> ...
> Also note sometimes when a function is pure/const it can be removed which is 
why it might act as 
> inlining.

This const/pure is definitely not my case.

I will try to write some example of non-inline. My project is huge, certainly
callgrind showed that there is stuff inlined, I will try to reproduce on
something smaller.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748

[Bug c++/19748] aggressive no-inline options still cause inlining

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-02-02 06:44 -------
(In reply to comment #4)
> This const/pure is definitely not my case.
How sure are you?

And are you sure that the options are being used to compile all the code.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pinskia at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748

[Bug c++/19748] aggressive no-inline options still cause inlining

[wilson at gcc dot gnu dot org]

------- Additional Comments From wilson at gcc dot gnu dot org  2005-02-11 23:11 -------
Try -fno-optimize-sibling-calls.

sibling-call (tail-call) optimizations can confuse anything that tries to
produce call graph info, and the end result will look similar to the result you
get with function inlining.  In both cases, a procedure frame gets optimized away.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748

[Bug c++/19748] aggressive no-inline options still cause inlining

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2005-05-13 11:53 -------
No feedback in 3 months.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19748