[Bug target/104915] New: Miss optimization for vec_setv8hi_0

[Bug target/104915] New: Miss optimization for vec_setv8hi_0

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104915

            Bug ID: 104915
           Summary: Miss optimization for vec_setv8hi_0
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: crazylht at gmail dot com
  Target Milestone: ---

#include<immintrin.h>
__m128i
foo (short* p)
{
    return _mm_set_epi32 (0, 0, 0, (unsigned short) ((*(__m16_u *)p)[0]));
}

__m128i
foo1 (short* p)
{
    return _mm_set_epi16 (0, 0, 0, 0, 0, 0, 0, (*(__m16_u *)p)[0]);
}

under avx512fp16, foo could generate vmovw instead of movzx + vmovd, without
avx512fp16 foo1 could generate movzx + movd instead of pxor + pinsrw.

[Bug target/104915] Miss optimization for vec_setv8hi_0

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104915

--- Comment #1 from Hongtao.liu <crazylht at gmail dot com> ---
As described in PR105066, pinsrw mem should be better than movzx + vmovd.

[Bug target/104915] Miss optimization for vec_setv8hi_0

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104915

--- Comment #2 from Hongtao.liu <crazylht at gmail dot com> ---
Created attachment 52705
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52705&action=edit
Patch pending for GCC13

[Bug target/104915] Miss optimization for vec_setv8hi_0

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104915

--- Comment #3 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by hongtao Liu <liuhongt@gcc.gnu.org>:

https://gcc.gnu.org/g:1e69bc810272f289e367570cf89d8c72af6124e4

commit r13-284-g1e69bc810272f289e367570cf89d8c72af6124e4
Author: liuhongt <hongtao.liu@intel.com>
Date:   Tue Mar 29 09:21:21 2022 +0800

    Optimize movzwl + vmovd/vmovq to vmovw.

    Similarly optimize movl + vmovq to vmovd.

    gcc/ChangeLog:

            PR target/104915
            * config/i386/sse.md (*vec_set<mode>_0_zero_extendhi): New
            pre_reload define_insn_and_split.
            (*vec_setv2di_0_zero_extendhi_1): Ditto.
            (*vec_set<mode>_0_zero_extendsi): Ditto.
            (*vec_setv2di_0_zero_extendsi_1): Ditto.
            (ssewvecmode): New mode attr.
            (ssewvecmodelower): Ditto.
            (ssepackmodelower): Ditto.

    gcc/testsuite/ChangeLog:

            * gcc.target/i386/pr104915-vmovd.c: New test.
            * gcc.target/i386/pr104915-vmovw.c: New test.

[Bug target/104915] Miss optimization for vec_setv8hi_0

[crazylht at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104915

Hongtao.liu <crazylht at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #4 from Hongtao.liu <crazylht at gmail dot com> ---
Fixed in GCC13.