[Bug c++/114134] New: Extra mov instructions for simple function compared with GCC13

[Bug c++/114134] New: Extra mov instructions for simple function compared with GCC13

[pilarlatiesa at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

            Bug ID: 114134
           Summary: Extra mov instructions for simple function compared
                    with GCC13
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pilarlatiesa at gmail dot com
  Target Milestone: ---

In the example below, the function `Key` has some extra (useless?) mov
instructions that are not generated with GCC 13.

$ cat borrar.cpp 

#include <cmath>

struct TVec3D { double x, y, z; };

struct TKey { int i, j, k; };

extern double const BinSize;

inline int Index(double const x)
  { return static_cast<int>(std::floor(static_cast<float>(x / BinSize + 1.0) -
1.0f)); };

TKey Key(TVec3D const &r)
  { return {Index(r.x), Index(r.y), Index(r.z)}; }


$ ./gcc-13/bin/g++ -O3 -march=skylake -fno-trapping-math -S borrar.cpp -o-
        .file   "borrar.cpp"
        .text
        .p2align 4
        .globl  _Z3KeyRK6TVec3D
        .type   _Z3KeyRK6TVec3D, @function
_Z3KeyRK6TVec3D:
.LFB993:
        .cfi_startproc
        vmovsd  BinSize(%rip), %xmm1
        vmovupd (%rdi), %xmm3
        vmovddup        .LC1(%rip), %xmm2
        vmovddup        %xmm1, %xmm0
        vdivpd  %xmm0, %xmm3, %xmm0
        vaddpd  %xmm2, %xmm0, %xmm0
        vmovq   .LC2(%rip), %xmm2
        vcvtpd2psx      %xmm0, %xmm0
        vaddps  %xmm2, %xmm0, %xmm0
        vroundps        $9, %xmm0, %xmm0
        vcvttps2dq      %xmm0, %xmm4
        vmovsd  16(%rdi), %xmm0
        vmovq   %xmm4, %rax
        vdivsd  %xmm1, %xmm0, %xmm0
        vaddsd  .LC1(%rip), %xmm0, %xmm0
        vcvtsd2ss       %xmm0, %xmm0, %xmm0
        vsubss  .LC3(%rip), %xmm0, %xmm0
        vroundss        $9, %xmm0, %xmm0, %xmm0
        vcvttss2sil     %xmm0, %edx
        movl    %edx, %edx
        ret
        .cfi_endproc
.LFE993:
        .size   _Z3KeyRK6TVec3D, .-_Z3KeyRK6TVec3D
        .section        .rodata.cst8,"aM",@progbits,8
        .align 8
.LC1:
        .long   0
        .long   1072693248
        .align 8
.LC2:
        .long   -1082130432
        .long   -1082130432
        .section        .rodata.cst4,"aM",@progbits,4
        .align 4
.LC3:
        .long   1065353216
        .ident  "GCC: (GNU) 13.1.0"
        .section        .note.GNU-stack,"",@progbits


$ ./gcc-14/bin/g++ -O3 -march=skylake -fno-trapping-math -S borrar.cpp -o-
        .file   "borrar.cpp"
        .text
        .p2align 4
        .globl  _Z3KeyRK6TVec3D
        .type   _Z3KeyRK6TVec3D, @function
_Z3KeyRK6TVec3D:
.LFB1032:
        .cfi_startproc
        vmovsd  BinSize(%rip), %xmm2
        vmovupd (%rdi), %xmm0
        vmovddup        %xmm2, %xmm1
        vdivpd  %xmm1, %xmm0, %xmm0
        vmovddup        .LC1(%rip), %xmm1
        vaddpd  %xmm1, %xmm0, %xmm0
        vmovq   .LC2(%rip), %xmm1
        vcvtpd2psx      %xmm0, %xmm0
        vaddps  %xmm1, %xmm0, %xmm0
        vroundps        $9, %xmm0, %xmm0
        vcvttps2dq      %xmm0, %xmm0
        vmovq   %xmm0, %rdx
        vmovsd  16(%rdi), %xmm0
        vdivsd  %xmm2, %xmm0, %xmm0
        vaddsd  .LC1(%rip), %xmm0, %xmm0
        vcvtsd2ss       %xmm0, %xmm0, %xmm0
        vsubss  .LC3(%rip), %xmm0, %xmm0
        vroundss        $9, %xmm0, %xmm0, %xmm0
        vcvttss2sil     %xmm0, %eax
        movl    %eax, %eax
        movq    %rax, %rdi
        movq    %rdx, %rax
        movq    %rdi, %rdx
        ret
        .cfi_endproc
.LFE1032:
        .size   _Z3KeyRK6TVec3D, .-_Z3KeyRK6TVec3D
        .section        .rodata.cst8,"aM",@progbits,8
        .align 8
.LC1:
        .long   0
        .long   1072693248
        .align 8
.LC2:
        .long   -1082130432
        .long   -1082130432
        .section        .rodata.cst4,"aM",@progbits,4
        .align 4
.LC3:
        .long   1065353216
        .ident  "GCC: (GNU) 14.0.0 20240112 (experimental)"
        .section        .note.GNU-stack,"",@progbits

[Bug target/114134] [14 Regression] Extra mov instructions for simple function compared with GCC13

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |14.0
            Summary|Extra mov instructions for  |[14 Regression] Extra mov
                   |simple function compared    |instructions for simple
                   |with GCC13                  |function compared with
                   |                            |GCC13

[Bug target/114134] [14 Regression] Extra mov instructions for simple function compared with GCC13

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2024-02-28

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Looks like the biggest issue is when expanding:
;; return D.26244;


Before:

(insn 34 33 35 (set (subreg:DI (reg:TI 107) 8)
        (reg:DI 108)) "/app/example.cpp":14:47 discrim 1 -1
     (nil))

(insn 35 34 36 (set (reg:TI 98 [ <retval> ])
        (reg:TI 107)) "/app/example.cpp":14:47 discrim 1 -1
     (nil))

After:
(insn 41 40 42 (set (reg:TI 130)
        (ior:TI (and:TI (reg:TI 130)
                (const_wide_int 0x0ffffffffffffffff))
            (ashift:TI (zero_extend:TI (reg:DI 131))
                (const_int 64 [0x40])))) "/app/example.cpp":14:47 discrim 1 -1
     (nil))

(insn 42 41 43 (set (reg:TI 114 [ <retval> ])
        (reg:TI 130)) "/app/example.cpp":14:47 discrim 1 -1
     (nil))

[Bug target/114134] [14 Regression] Extra mov instructions for simple function compared with GCC13

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |sayle at gcc dot gnu.org
             Target|X86_64                      |x86_64-*-*

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
I guess the testcase can be simplified to just show the return value handling
issue.

[Bug target/114134] [14 Regression] Extra mov instructions for simple function compared with GCC13

[pilarlatiesa at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

--- Comment #3 from Pilar Latiesa <pilarlatiesa at gmail dot com> ---
(In reply to Richard Biener from comment #2)
> I guess the testcase can be simplified to just show the return value
> handling issue.

I think this suffices:

struct TVec3D { double x, y, z; };

struct TKey { int i, j, k; };

TKey Key(TVec3D const &r)
  { return {int(r.x), int(r.y), int(r.z)}; }

[Bug target/114134] [14 Regression] Extra mov instructions for simple function compared with GCC13 since r14-2386

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[14 Regression] Extra mov   |[14 Regression] Extra mov
                   |instructions for simple     |instructions for simple
                   |function compared with      |function compared with
                   |GCC13                       |GCC13 since r14-2386
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Started with r14-2386-gbdf2737cda53a83332db1a1a021653447b05a7e7

[Bug target/114134] [14 Regression] Extra mov instructions for simple function compared with GCC13 since r14-2386

[pilarlatiesa at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

--- Comment #5 from Pilar Latiesa <pilarlatiesa at gmail dot com> ---
Another testcase:

struct TKey { int i, j, k, w; };

TKey Key(int x)
  { return {x, 0, x, 0}; }

[Bug target/114134] [14 Regression] Extra mov instructions for simple function compared with GCC13 since r14-2386

[law at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

Jeffrey A. Law <law at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P2
                 CC|                            |law at gcc dot gnu.org

[Bug target/114134] [14/15 Regression] Extra mov instructions for simple function compared with GCC13 since r14-2386

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114134

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|14.0                        |14.2

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 14.1 is being released, retargeting bugs to GCC 14.2.