[Bug target/126784] New: Sub-optimal code sequence for __builtin_convertvector

[hjl.tools at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=126784

            Bug ID: 126784
           Summary: Sub-optimal code sequence for __builtin_convertvector
           Product: gcc
           Version: 17.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hjl.tools at gmail dot com
                CC: liuhongt at gcc dot gnu.org
  Target Milestone: ---
            Target: x86

GCC generates the optimal code sequence for this:

[hjl@gnu-zen4-1 vect-1]$ cat x.C
typedef int  v2si __attribute__((vector_size (8)));
typedef long long v2di __attribute__((vector_size (16)));
v2si
f (v2si a, v2si b)
{
  v2di x = __builtin_convertvector (a, v2di);
  v2di y = __builtin_convertvector (b, v2di);
  v2di z = x < y ? x : y;
  return __builtin_convertvector (z, v2si);
}
[hjl@gnu-zen4-1 vect-1]$ make x.s
/export/build/gnu/tools-build/gcc-gitlab-debug/build-x86_64-linux/gcc/xgcc
-B/export/build/gnu/tools-build/gcc-gitlab-debug/build-x86_64-linux/gcc/ -O2
-msse4 -S x.C
[hjl@gnu-zen4-1 vect-1]$ cat x.s
        .file   "x.C"
        .text
        .p2align 4
        .globl  _Z1fDv2_iS_
        .type   _Z1fDv2_iS_, @function
_Z1fDv2_iS_:
.LFB0:
        .cfi_startproc
        pminsd  %xmm1, %xmm0
        ret
        .cfi_endproc
.LFE0:
        .size   _Z1fDv2_iS_, .-_Z1fDv2_iS_
        .ident  "GCC: (GNU) 17.0.0 20260810 (experimental)"
        .section        .note.GNU-stack,"",@progbits
[hjl@gnu-zen4-1 vect-1]$ 

but generates sub-optimal code sequence for slightly modified source:

[hjl@gnu-zen4-1 vect-1]$ cat y.C
typedef int v2si __attribute__((vector_size (8)));
typedef long long v2di __attribute__((vector_size (16)));

v2si
f (v2si a, v2si b, v2di *p)
{
  v2di x = __builtin_convertvector (a, v2di);
  v2di y = __builtin_convertvector (b, v2di);
  v2di z = x < y ? x : y;
  *p = z;
  return __builtin_convertvector (z, v2si);
}
[hjl@gnu-zen4-1 vect-1]$ make y.s
/export/build/gnu/tools-build/gcc-gitlab-debug/build-x86_64-linux/gcc/xgcc
-B/export/build/gnu/tools-build/gcc-gitlab-debug/build-x86_64-linux/gcc/ -O2
-msse4 -S y.C
[hjl@gnu-zen4-1 vect-1]$ cat y.s
        .file   "y.C"
        .text
        .p2align 4
        .globl  _Z1fDv2_iS_PDv2_x
        .type   _Z1fDv2_iS_PDv2_x, @function
_Z1fDv2_iS_PDv2_x:
.LFB0:
        .cfi_startproc
        movdqa  %xmm0, %xmm2
        pmovsxdq        %xmm1, %xmm1
        pmovsxdq        %xmm2, %xmm2
        movdqa  %xmm2, %xmm0
        pcmpgtq %xmm1, %xmm0
        pand    %xmm0, %xmm1
        pandn   %xmm2, %xmm0
        por     %xmm1, %xmm0
        movaps  %xmm0, (%rdi)
        shufps  $232, %xmm0, %xmm0
        ret
        .cfi_endproc
.LFE0:
        .size   _Z1fDv2_iS_PDv2_x, .-_Z1fDv2_iS_PDv2_x
        .ident  "GCC: (GNU) 17.0.0 20260810 (experimental)"
        .section        .note.GNU-stack,"",@progbits
[hjl@gnu-zen4-1 vect-1]$