[Bug tree-optimization/96966] New: redundant memcpy not eliminated after pointer subtraction

[Bug tree-optimization/96966] New: redundant memcpy not eliminated after pointer subtraction

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

            Bug ID: 96966
           Summary: redundant memcpy not eliminated after pointer
                    subtraction
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: msebor at gcc dot gnu.org
  Target Milestone: ---

The second call to memcpy and mempcpy, respectively, is redundant in each of
the functions below and can be eliminated.  GCC doesn't notice the redundancy
and emits both pairs of copies.  Clang emits just one call to memcpy but it too
fails to eliminate the redundant call to mempcpy.

$ cat z.c && gcc -O2 -S -o /dev/stdout z.c
extern char a[32];

void f (const void *s)
{
  char *p = (char*)__builtin_memcpy (a, s, 16) + 16;
  __builtin_memcpy (p - 16, s, 16);
}


void g (const void *s)
{
  char *p = (char*)__builtin_mempcpy (a, s, 16);
  __builtin_mempcpy (p - 16, s, 16);
}
        .file   "z.c"
        .text
        .p2align 4
        .globl  f
        .type   f, @function
f:
.LFB0:
        .cfi_startproc
        movdqu  (%rdi), %xmm0
        movups  %xmm0, a(%rip)
        movdqu  (%rdi), %xmm1
        movups  %xmm1, a(%rip)
        ret
        .cfi_endproc
.LFE0:
        .size   f, .-f
        .p2align 4
        .globl  g
        .type   g, @function
g:
.LFB3:
        .cfi_startproc
        movdqu  (%rdi), %xmm0
        movups  %xmm0, a(%rip)
        movdqu  (%rdi), %xmm1
        movups  %xmm1, a(%rip)
        ret
        .cfi_endproc
.LFE3:
        .size   g, .-g
        .ident  "GCC: (GNU) 11.0.0 20200902 (experimental)"
        .section        .note.GNU-stack,"",@progbits

[Bug tree-optimization/96966] [8/9/10/11 Regression] redundant memcpy not eliminated after pointer subtraction

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

Martin Sebor <msebor at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|redundant memcpy not        |[8/9/10/11 Regression]
                   |eliminated after pointer    |redundant memcpy not
                   |subtraction                 |eliminated after pointer
                   |                            |subtraction
      Known to work|                            |8.1.0, 8.2.0
      Known to fail|                            |10.2.0, 11.0, 7.3.0, 8.3.0,
                   |                            |9.2.0
           Keywords|                            |missed-optimization

--- Comment #1 from Martin Sebor <msebor at gcc dot gnu.org> ---
According to Godbolt, GCC 8.1 and 8.2 emit optimal code for both functions but
GCC 8.3 emits the less optimal code for f and has g jump to it.  Starting with
10.1, GCC emits the same suboptimal code for both functions.

[Bug tree-optimization/96966] [8/9/10/11 Regression] redundant memcpy not eliminated after pointer subtraction

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2020-09-08
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
           Keywords|                            |alias
   Target Milestone|---                         |8.5

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Martin Sebor from comment #1)
> According to Godbolt, GCC 8.1 and 8.2 emit optimal code for both functions
> but GCC 8.3 emits the less optimal code for f and has g jump to it. 
> Starting with 10.1, GCC emits the same suboptimal code for both functions.

This is likely "caused" by 08dfb1d682a707f7319aafec28edda424395dae5, aka
the fix for PR91108 which was also backported.  In the IL

  <bb 2> :
  _5 = MEM <__int128 unsigned> [(char * {ref-all})s_4(D)];
  MEM <__int128 unsigned> [(char * {ref-all})&a] = _5;
  _8 = MEM <__int128 unsigned> [(char * {ref-all})s_4(D)];
  MEM <__int128 unsigned> [(char * {ref-all})&a] = _8;
  return;

we lost the information that MEM <__int128 unsigned> [(char * {ref-all})s_4(D)]
and MEM <__int128 unsigned> [(char * {ref-all})&a] do not partially overlap.
The memcpy call guaranteed that.  The way the aliasing code rules out
partial overlap is by using alignment which doesn't help us here.

That it worked in GCC 8.[12] was due to bad code in VN that ignored
the possibility of a partial overlap here.

We eventually could lower memcpy (a, s, 16) to load + store with noting
they are independent using MR_DEPENDENCE_CLIQUE/BASE but this may cause
depleting of the clique resource on artificial testcases quickly
(we only have 16bits for clique).  Shifting bit allocation between
clique and base might be a possibility there, but at least clique
overflow mitigation would need to be put in place.

[Bug tree-optimization/96966] [8/9/10/11 Regression] redundant memcpy not eliminated after pointer subtraction

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2

[Bug tree-optimization/96966] [8/9/10/11 Regression] redundant memcpy not eliminated after pointer subtraction

[msebor at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

--- Comment #3 from Martin Sebor <msebor at gcc dot gnu.org> ---
No change in GCC 11.

[Bug tree-optimization/96966] [8/9/10/11 Regression] redundant memcpy not eliminated after pointer subtraction

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Even
extern char a[32];

void f (const void *s)
{
  char *p = (char*)__builtin_memcpy (a, s, 16);
  __builtin_memcpy (p, s, 16);
}


void g (const void *s)
{
  __builtin_memcpy (a, s, 16);
  __builtin_memcpy (a, s, 16);
}
used to be optimized just in 8.1/8.2 and not in earlier or later GCC versions.
Perhaps delaying the lowering of memcpy a tiny bit and trying to optimize it
when it is still not lowered?

[Bug tree-optimization/96966] [8/9/10/11 Regression] redundant memcpy not eliminated after pointer subtraction

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Jakub Jelinek from comment #4)
> Even
> extern char a[32];
> 
> void f (const void *s)
> {
>   char *p = (char*)__builtin_memcpy (a, s, 16);
>   __builtin_memcpy (p, s, 16);
> }
> 
> 
> void g (const void *s)
> {
>   __builtin_memcpy (a, s, 16);
>   __builtin_memcpy (a, s, 16);
> }
> used to be optimized just in 8.1/8.2 and not in earlier or later GCC
> versions.
> Perhaps delaying the lowering of memcpy a tiny bit and trying to optimize it
> when it is still not lowered?

lowering early is quite an important thing since it enables better initial
into-SSA rewriting and early inline costing.  Note even w/o lowering FRE
would not optimize this.  Even the strlen pass doesn't:

extern char a[32];
void __GIMPLE (ssa,startwith("fre1")) g (const void *s)
 {
__BB(2):
   __builtin_memcpy (&a[0], s_1(D), _Literal (__SIZE_TYPE__) 16);
   __builtin_memcpy (&a[0], s_1(D), _Literal (__SIZE_TYPE__) 16);
   return;
 }

has this survive until .fab if you do -O2 -fno-tree-forwprop -fno-tree-vrp

[Bug tree-optimization/96966] [9/10/11/12 Regression] redundant memcpy not eliminated after pointer subtraction

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|8.5                         |9.4

--- Comment #6 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 8 branch is being closed.

[Bug tree-optimization/96966] [9/10/11/12 Regression] redundant memcpy not eliminated after pointer subtraction

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|9.4                         |9.5

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 9.4 is being released, retargeting bugs to GCC 9.5.

[Bug tree-optimization/96966] [10/11/12/13 Regression] redundant memcpy not eliminated after pointer subtraction

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|9.5                         |10.4

--- Comment #8 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 9 branch is being closed

[Bug tree-optimization/96966] [10/11/12/13 Regression] redundant memcpy not eliminated after pointer subtraction

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.4                        |10.5

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 10.4 is being released, retargeting bugs to GCC 10.5.

[Bug tree-optimization/96966] [11/12/13/14 Regression] redundant memcpy not eliminated after pointer subtraction

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96966

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|10.5                        |11.5

--- Comment #10 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 10 branch is being closed.