[Bug gcov-profile/82527] Branch probabilities does not match on optimized switch cases

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82527

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |INVALID

--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
Thanks for the issue. Note that order of branches which go from switch
statement to corresponding case labels is not ordered.
With new GCC 8.1, we'll come up with '-w' option. Then you'll see the
connections (edges):

$ rm *.gcda && gcc  pr82527.c --coverage && ./a.out && gcov -abw pr82527.c &&
cat pr82527.c.gcov
0dddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddFile
'pr82527.c'
Lines executed:83.33% of 12
Branches executed:100.00% of 5
Taken at least once:80.00% of 5
No calls
Creating 'pr82527.c.gcov'

        -:    0:Source:pr82527.c
        -:    0:Graph:pr82527.gcno
        -:    0:Data:pr82527.gcda
        -:    0:Runs:1
        -:    0:Programs:1
        -:    1:#include <stdio.h>
        -:    2:int
function main called 1 returned 100% blocks executed 91%
        1:    3:main ()
        -:    4:{
      101:    5:  for (int i = 0; i < 100; i++)
        1:    5-block  0 (BB 2)
      100:    5-block  1 (BB 9)
      101:    5-block  2 (BB 10)
branch  0 taken 99% (BB 3)
branch  1 taken 1% (fallthrough) (BB 11)
        -:    6:    {
      100:    7:      switch (i)
      100:    7-block  0 (BB 3)
branch  0 taken 1% (BB 4)
branch  1 taken 0% (BB 6)
branch  2 taken 99% (BB 7)
        -:    8:        {
        1:    9:        case 0:
        1:   10:          printf ("0");
        1:   10-block  0 (BB 4)
call    0 returned 100%
        1:   11:          break;
    #####:   12:        case 9999:
    #####:   13:          break;
    $$$$$:   13-block  0 (BB 6)
       99:   14:        default:
       99:   15:          printf ("d");
       99:   15-block  0 (BB 7)
call    0 returned 100%
       99:   16:          break;
        -:   17:        }
        -:   18:    }
        1:   19:  return 0;
        1:   19-block  0 (BB 11)
        -:   20:}

$ rm *.gcda && gcc -O3 pr82527.c --coverage && ./a.out && gcov -abw pr82527.c
&& cat pr82527.c.gcov
0dddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddFile
'pr82527.c'
Lines executed:100.00% of 9
Branches executed:100.00% of 4
Taken at least once:75.00% of 4
No calls
Creating 'pr82527.c.gcov'

        -:    0:Source:pr82527.c
        -:    0:Graph:pr82527.gcno
        -:    0:Data:pr82527.gcda
        -:    0:Runs:1
        -:    0:Programs:1
        -:    1:#include <stdio.h>
        -:    2:int
function main called 1 returned 100% blocks executed 100%
        1:    3:main ()
        -:    4:{
      101:    5:  for (int i = 0; i < 100; i++)
        1:    5-block  0 (BB 2)
      100:    5-block  1 (BB 8)
      101:    5-block  2 (BB 9)
branch  0 taken 99% (BB 3)
branch  1 taken 1% (fallthrough) (BB 10)
        -:    6:    {
      100:    7:      switch (i)
      100:    7-block  0 (BB 3)
branch  0 taken 1% (BB 4)
branch  1 taken 99% (BB 6)
branch  2 taken 0% (BB 8)
        -:    8:        {
        1:    9:        case 0:
        1:   10:          printf ("0");
        1:   10-block  0 (BB 4)
call    0 returned 100%
        1:   11:          break;
        -:   12:        case 9999:
        -:   13:          break;
       99:   14:        default:
       99:   15:          printf ("d");
       99:   15-block  0 (BB 6)
call    0 returned 100%
       99:   16:          break;
        -:   17:        }
        -:   18:    }
        -:   19:  return 0;
        -:   20:}

And you're right w/ -O3 it goes to a different BB and order is changed.
Hope it will help you? I'm closing it as invalid.