[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139 Andrew Pinski changed: What|Removed |Added CC||eric-bugs at omnifarious dot org --- Comment #9 from Andrew Pinski --- *** Bug 108361 has been marked as a duplicate of this bug. ***
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139 Andrew Pinski changed: What|Removed |Added CC||witold.baryluk+gcc at gmail dot co ||m --- Comment #8 from Andrew Pinski --- *** Bug 105360 has been marked as a duplicate of this bug. ***
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139 Andrew Pinski changed: What|Removed |Added Severity|normal |enhancement
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139
Patrick Palka changed:
What|Removed |Added
CC||ppalka at gcc dot gnu.org
--- Comment #7 from Patrick Palka ---
Here's a simple C++ example using std::function that illustrates the same
thing. The call to h below is optimized away, but two now-unused static member
functions _M_invoke and _M_manager remain in the optimized code.
#include
int
m ()
{
std::function h = [] (int i) { return i - 1; };
return h (5);
}
-fdump-tree-optimized:
;; Function std::_Function_handler >::_M_invoke
(_ZNSt17_Function_handlerIFiiEZ1mvEUliE_E9_M_invokeERKSt9_Any_dataOi,
funcdef_no=1738, decl_uid=33290, cgraph_uid=651, symbol_order=654)
std::_Function_handler >::_M_invoke (const union
_Any_data & {ref-all} __functor, int & __args#0)
{
...
}
;; Function std::_Function_base::_Base_manager >::_M_manager
(_ZNSt14_Function_base13_Base_managerIZ1mvEUliE_E10_M_managerERSt9_Any_dataRKS3_St18_Manager_operation,
funcdef_no=1739, decl_uid=33254, cgraph_uid=652, symbol_order=655)
Removing basic block 7
Removing basic block 9
std::_Function_base::_Base_manager >::_M_manager (union
_Any_data & {ref-all} __dest, const union _Any_data & {ref-all} __source,
_Manager_operation __op)
{
...
}
;; Function m (_Z1mv, funcdef_no=1392, decl_uid=31150, cgraph_uid=311,
symbol_order=314)
m ()
{
[local count: 1073741824]:
return 4;
}
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139
--- Comment #6 from Marc Nieper-Wißkirchen ---
Interestingly, the above code, slightly modified, exhibits another problem the
optimizer has:
void foo ();
static void c (void (*w) (void))
{
w ();
}
static void f ()
{
foo ();
}
static void *x =
void g ()
{
c (x);
}
The only difference consists in that f is not completely trivial but calls some
external function. In this case, the code of f is not inlined in g, and we get
the non-optimal code (in x86-64 assembly):
f:
xorl%eax, %eax
jmp foo
g:
jmp f
I'm not sure whether I should file in another bug for this example.
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139
--- Comment #5 from Marc Nieper-Wißkirchen ---
Here is a much simpler example, which exhibits the same behavior:
static void c (void (*w) (void))
{
w ();
}
static void f ()
{
}
static void *x =
void g ()
{
c (x);
}
The compiler optimizes g to an empty procedure and removes that static variable
x from the output. However, it outputs the unused, static function f.
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139 --- Comment #4 from Marc Nieper-Wißkirchen --- P.S.: This issue showed up when I tried to analyze why no optimization is happening in bug #89152.
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139 --- Comment #3 from David Malcolm --- (In reply to David Malcolm from comment #2) > happens when we're already running cgraph_node::expand on "h" ~~ "when we've already run", I meant to say
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139
David Malcolm changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed||2019-02-01
CC||dmalcolm at gcc dot gnu.org
Ever confirmed|0 |1
--- Comment #2 from David Malcolm ---
Indeed, the last elimination is happening too late.
At -O3:
.040t.tailr1 converts the tail-recursion in "g" into a loop
.076i.inline inlines "g" into "foo", but an initialization of a temporary with
"h" remains in "foo"
.103t.dce2 eliminates that last use of "h".
With the "a * 2" -> "a" variant at -O3,
.025t.einline inlines "g" into "foo", converting the tail recursion into a
loop
.033t.fre1 eliminates the usage of "h"
cgraphunit.c:expand_all_functions happens at .090t ("fixup_cfg" and
siblings/descendants), so the elimination of the last "h" in .103t.dce2 happens
when we're already running cgraph_node::expand on "h" (whereas in the einline
variant, "h" has already been eliminated before
cgraphunit.c:expand_all_functions).
[Bug ipa/89139] GCC emits code for static functions that aren't used by the optimized code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89139 Richard Biener changed: What|Removed |Added Keywords||missed-optimization CC||marxin at gcc dot gnu.org Component|tree-optimization |ipa Version|unknown |9.0 --- Comment #1 from Richard Biener --- It's likely eliminating the last use too late.
