--- Comment #9 from pinskia at earth dot geop dot uc dot edu 2006-10-23
21:52 ---
Subject: Re: Gcc gives invalid warning about unitialized variable
>
>
>
> --- Comment #8 from mbligh at mbligh dot org 2006-10-23 21:50 ---
> > Initialize the variable and forget about ineffi
>
>
>
> --- Comment #8 from mbligh at mbligh dot org 2006-10-23 21:50 ---
> > Initialize the variable and forget about inefficiency. Again this is fixed
> > for
> > 4.2.0, the warning is only because __builtin_expect gets in the way of
> > figuring
> > out if the variable is used uni
--- Comment #8 from mbligh at mbligh dot org 2006-10-23 21:50 ---
> Initialize the variable and forget about inefficiency. Again this is fixed
> for
> 4.2.0, the warning is only because __builtin_expect gets in the way of
> figuring
> out if the variable is used uninitialized, nothing
--- Comment #7 from pinskia at gcc dot gnu dot org 2006-10-23 21:45 ---
(In reply to comment #6)
> Re the non-inlined functions ...
> Have been discussing this with Andrew Morton.
>
> what do we do then? Adding dead code to fix the fact that gcc can't see into
> other functions is incor
--- Comment #6 from mbligh at mbligh dot org 2006-10-23 21:42 ---
Re the non-inlined functions ...
Have been discussing this with Andrew Morton.
what do we do then? Adding dead code to fix the fact that gcc can't see into
other functions is incorrect and inefficient.
Having lots of com
--- Comment #5 from pinskia at gcc dot gnu dot org 2006-10-23 21:30 ---
(In reply to comment #4)
> And indeed, if I remove that unlikely(), it does work.
>
> There's another set of these where the called initializer is not inlined,
> ie:
>
> int x;
> initializer(&x);
If initializer is
--- Comment #4 from mbligh at mbligh dot org 2006-10-23 21:19 ---
And indeed, if I remove that unlikely(), it does work.
There's another set of these where the called initializer is not inlined,
ie:
int x;
initializer(&x);
Which it seems totally blind to, even if the initializer does
--- Comment #3 from mbligh at mbligh dot org 2006-10-23 21:08 ---
Yeah, is builtin_expect ;-(
#define likely(x) __builtin_expect(!!(x), 1)
#define unlikely(x) __builtin_expect(!!(x), 0)
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29574
--- Comment #2 from pinskia at gcc dot gnu dot org 2006-10-23 20:59 ---
If it is defined using __builtin_expect, then this is most likely PR 21513
which was fixed for 4.2.0, it is hard to fix correctly for 4.1.0 and 4.0.0
without removing all of loop.c.
Can you attach the preprocessed s
--- Comment #1 from pinskia at gcc dot gnu dot org 2006-10-23 20:57 ---
How is unlikely defined?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29574
10 matches
Mail list logo
