[Bug ipa/100100] New: missed optimization for dead code elimination at -O3 (vs. -O1, -Os, -O2)

[Bug ipa/100100] New: missed optimization for dead code elimination at -O3 (vs. -O1, -Os, -O2)

[zhendong.su at inf dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100100

            Bug ID: 100100
           Summary: missed optimization for dead code elimination at -O3
                    (vs. -O1, -Os, -O2)
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: ipa
          Assignee: unassigned at gcc dot gnu.org
          Reporter: zhendong.su at inf dot ethz.ch
                CC: marxin at gcc dot gnu.org
  Target Milestone: ---

[538] % gcctk -v
Using built-in specs.
COLLECT_GCC=gcctk
COLLECT_LTO_WRAPPER=/local/suz-local/software/local/gcc-trunk/libexec/gcc/x86_64-pc-linux-gnu/11.0.1/lto-wrapper
Target: x86_64-pc-linux-gnu
Configured with: ../gcc-trunk/configure --disable-bootstrap
--prefix=/local/suz-local/software/local/gcc-trunk --enable-languages=c,c++
--disable-werror --enable-multilib --with-system-zlib
Thread model: posix
Supported LTO compression algorithms: zlib
gcc version 11.0.1 20210415 (experimental) [master revision
4dd9e1c541e:7315804b0a0:b5f644a98b3f3543d3a8d2dfea7785c22879013f] (GCC) 
[539] % 
[539] % gcctk -O1 -S -o O1.s small.c
[540] % gcctk -O3 -S -o O3.s small.c
[541] % 
[541] % wc O1.s O3.s
  79  175 1126 O1.s
  98  196 1466 O3.s
 177  371 2592 total
[542] % 
[542] % grep foo O1.s
[543] % grep foo O3.s
        jmp     foo
[544] % 
[544] % cat small.c
extern void foo(void);
static int a, d, e, h, o, p;
static short b, c, f, *i = &b;
static char g;
static void l() {
  d = 0;
  for (; d < 4; d++)
    for (; f; f++)
      c = 0;
  for (; c; c++)
    ;
}
static void k(unsigned short q) {
  l();
  for (; g; g++)
    h = e == p;
}
static void j(int q) {
  k(q);
  unsigned char m = q;
  o = m;
  if (m)
    foo();
}
void n() {
  j(a);
  *i = 0;
}
int main() {
  n();
  return 0;
}

[Bug ipa/100100] missed optimization for dead code elimination at -O3 (vs. -O1, -Os, -O2)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100100

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jamborm at gcc dot gnu.org
            Version|unknown                     |11.0
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |missed-optimization
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2021-04-15

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Different inlining somehow causes different CCP:

-;; Function n (n, funcdef_no=3, decl_uid=1980, cgraph_uid=4, symbol_order=14)
+;; Function j (j, funcdef_no=2, decl_uid=1977, cgraph_uid=3, symbol_order=13)

 Adding destination of edge (0 -> 2) to worklist

 Simulating block 2

 Visiting statement:
-a.0_1 = a;
-which is likely CONSTANT
-Lattice value changed to CONSTANT 0.  Adding SSA edges to worklist.
-marking stmt to be not simulated again
-
-Visiting statement:
-_6 = a.0_1 & 65535;
+_8 = q_2(D) & 65535;
 which is likely CONSTANT
-Match-and-simplified a.0_1 & 65535 to 0
-Lattice value changed to CONSTANT 0.  Adding SSA edges to worklist.
+Lattice value changed to CONSTANT 0x0 (0xffff).  Adding SSA edges to worklist.

so we know 'a' is zero (after IPA) but at -O3 a is passed as parameter
(but IPA CP didn't figure out its constant value).

Still odd that the inlining decision is different.