[Bug tree-optimization/97856] New: Missed optimization: repeated call

[Bug tree-optimization/97856] New: Missed optimization: repeated call

[miguel.ojeda.sandonis at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97856

            Bug ID: 97856
           Summary: Missed optimization: repeated call
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: miguel.ojeda.sandonis at gmail dot com
  Target Milestone: ---

For code such as:

    void f(int);
    void g(int v) {
        switch (v) {
            case 1: f(1);
            case 2: f(2);
            case 3: f(3);
            default: f(4);
        }
    }

GCC generates for e.g. x86_64:

    g(int):
            subq    $8, %rsp
            cmpl    $2, %edi
            je      .L2
            cmpl    $3, %edi
            je      .L3
            cmpl    $1, %edi
            je      .L13
            movl    $4, %edi
            addq    $8, %rsp
            jmp     f(int)
    .L13:
            call    f(int)
    .L2:
            movl    $2, %edi
            call    f(int)
    .L3:
            movl    $3, %edi
            call    f(int)
            movl    $4, %edi
            addq    $8, %rsp
            jmp     f(int)

Repeating the `f(4);` call can be avoided by reordering the cases, reducing
code size.

[Bug rtl-optimization/97856] Missed optimization: repeated call

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97856

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
            Version|unknown                     |11.0
          Component|tree-optimization           |rtl-optimization
     Ever confirmed|0                           |1
                 CC|                            |hubicka at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2020-11-17

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.  basic-block reordering decides to duplicate the block:

Redirecting fallthru edge 7->8 to 9
Duplicated bb 8 (created bb 9)
Final order:
2 3 4 8 5 6 7 9
Reordered sequence:
 2 bb 2
 3 bb 3
 4 bb 4
 5 bb 8
 6 bb 5
 7 bb 6
 8 bb 7
 9 duplicate of 8

since we decide to put the default: handling on the fallthru edge of the
condition chain we duplicate it to the tail as well.  IIRC expansion
always does it this way and BB reorder simply avoids one jump.  With -Os
we flip the last jump and avoid the duplication.

[Bug rtl-optimization/97856] Missed optimization: repeated call

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97856

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #1)
> Confirmed.  basic-block reordering decides to duplicate the block:

Yes there are a few other bugs where we like to duplicate the return block I
have seen too.

[Bug rtl-optimization/97856] Missed optimization: repeated call

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97856

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement