[Bug target/102737] New: [x86] Failure to optimize out bad register usage involving int->double conversion

[Bug target/102737] New: [x86] Failure to optimize out bad register usage involving int->double conversion

[gabravier at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102737

            Bug ID: 102737
           Summary: [x86] Failure to optimize out bad register usage
                    involving int->double conversion
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gabravier at gmail dot com
  Target Milestone: ---

double foo(int s, double a)
{
  return s + a;
}

With -O3, GCC outputs this on AMD64:

foo(int, double):
        movapd  xmm1, xmm0
        pxor    xmm0, xmm0
        cvtsi2sd        xmm0, edi
        addsd   xmm0, xmm1
        ret

LLVM outputs this:

foo(int, double):
        cvtsi2sd        xmm1, edi
        addsd   xmm0, xmm1
        ret

The movapd in GCC's version can be optimized out.

[Bug target/102737] [9/10/11/12 Regression] extra mov with int->double conversion and addition (incoming arguments and return)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102737

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |9.5
      Known to fail|                            |10.1.0, 9.1.0
           Severity|normal                      |enhancement
      Known to work|                            |5.1.0, 8.1.0, 8.5.0
            Summary|[x86] Failure to optimize   |[9/10/11/12 Regression]
                   |out bad register usage      |extra mov with int->double
                   |involving int->double       |conversion and addition
                   |conversion                  |(incoming arguments and
                   |                            |return)

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
GCC 8.5.0 (and before) would do:
        pxor    %xmm1, %xmm1
        cvtsi2sd        %edi, %xmm1
        addsd   %xmm1, %xmm0
        ret

So this is a regression.

Even with AVX, gcc has an extra move even if inputs don't have to be tied to
the output:
        vmovsd  %xmm0, %xmm0, %xmm1
        vxorps  %xmm0, %xmm0, %xmm0
        vcvtsi2sdl      %edi, %xmm0, %xmm0
        vaddsd  %xmm1, %xmm0, %xmm0
        ret

[Bug target/102737] [9/10/11/12 Regression] extra mov with int->double conversion and addition (incoming arguments and return)

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102737

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-10-13
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
           Severity|enhancement                 |minor

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Note I think this might not be an important regression either because I suspect
it has to do with function arguments so this would be a micro-benchmark really
rather than hitting real code.

[Bug target/102737] [9/10/11/12 Regression] extra mov with int->double conversion and addition (incoming arguments and return)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102737

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

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

[Bug target/102737] [10/11/12/13 Regression] extra mov with int->double conversion and addition (incoming arguments and return)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102737

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

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

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 9 branch is being closed

[Bug target/102737] [10/11/12/13 Regression] extra mov with int->double conversion and addition (incoming arguments and return)

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102737

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

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

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
GCC 10.4 is being released, retargeting bugs to GCC 10.5.

[Bug target/102737] [11/12/13/14 Regression] extra mov with int->double conversion and addition (incoming arguments and return)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102737

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

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

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 10 branch is being closed.