[Bug tree-optimization/17217] not removing removal of nested structs

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17217

Andrew Pinski  changed:

   What|Removed |Added

 Status|UNCONFIRMED |NEW
   Last reconfirmed|2005-01-25 03:37:17 |2021-09-03
 Ever confirmed|0   |1

--- Comment #8 from Andrew Pinski  ---
So where is a better testcase (which is not depedent on how gimplifier works):
int h(int *a);
int f(int i, int j)
{
  int t = i;
  int t1 = j;
  int g()
  {
int t11 = t1;
int t2 = h();
return t2 + t11;
  }
  int t3 = g();
  return t3 + t1;
}

 CUT ---
On the gimple level we get:
  _3 = __builtin_dwarf_cfa (0);
  FRAME.1.FRAME_BASE.PARENT = _3;
  FRAME.1.t = i_5(D);
  FRAME.1.t1 = j_7(D);
  t2_11 = h ();
  _12 = j_7(D) + t2_11;
  _9 = FRAME.1.t1;
  _10 = _9 + _12;

We should know that FRAME.1 is special in that h can only escape t field of
FRAME.1 and not the other fields of FRAME.1 too.

[Bug tree-optimization/17217] not removing removal of nested structs

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17217

Andrew Pinski  changed:

   What|Removed |Added

 Status|WAITING |UNCONFIRMED
 Ever confirmed|1   |0

--- Comment #7 from Andrew Pinski  ---
(In reply to Steven Bosscher from comment #6)
> What code is expected in this really old bug? Trunk x86-64 today:
> __attribute__((noinline))

g1 is not to be noinline for the nested testcase.

The un-nested testcase is actually should not be optimized.
The nested function testcase shows the issue.
For the original testcase:
int h(int *a);
int f(int i, int j)
{
  int t = i;
  int t1 = j;
  int g()
  {
 return h() + t1;
  }
  return g() + t1;
}

We should optimize this to the same as:
int h(int *a);
int f(int i, int j)
{
  return h() + 2 * j;
}

[Bug tree-optimization/17217] not removing removal of nested structs

[steven at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=17217

Steven Bosscher  changed:

   What|Removed |Added

 Status|REOPENED|WAITING

--- Comment #6 from Steven Bosscher  ---
What code is expected in this really old bug? Trunk x86-64 today:

int h(int *a);
struct G
{
int t;
int t1;
};
static int g1(struct G*);
int f1(int i, int j)
{
  struct G nestedf1={i,j};
  return g1()+nestedf1.t1;
}
__attribute__((noinline))
static int g1(struct G *nestedf1)
{
  return h(>t)+ nestedf1->t1;
}



g1(G*):
pushq   %rbx
movq%rdi, %rbx
callh(int*)
addl4(%rbx), %eax
popq%rbx
ret
f1(int, int):
subq$24, %rsp
movl%edi, 8(%rsp)
leaq8(%rsp), %rdi
movl%esi, 12(%rsp)
callg1(G*)
addl12(%rsp), %eax
addq$24, %rsp
ret

[Bug tree-optimization/17217] not removing removal of nested structs

[pinskia at gcc dot gnu dot org]

--- Additional Comments From pinskia at gcc dot gnu dot org  2005-04-16 
17:37 ---
Fixed in 4.1.0 and above by Daniel's aliasing improvements.

-- 
   What|Removed |Added

 Status|NEW |RESOLVED
 Resolution||FIXED
   Target Milestone|--- |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17217