[Bug target/101900] New: -mtune-ctrl=use_vector_fp_converts,use_vector_converts doesn't work with AVX

[Bug target/101900] New: -mtune-ctrl=use_vector_fp_converts,use_vector_converts doesn't work with AVX

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101900

            Bug ID: 101900
           Summary: -mtune-ctrl=use_vector_fp_converts,use_vector_converts
                    doesn't work with AVX
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hjl.tools at gmail dot com
                CC: crazylht at gmail dot com
  Target Milestone: ---
            Target: i386,x86-64

[hjl@gnu-tgl-2 pr87007]$ cat x.c
extern float f;
extern double d;
extern int i;

void
foo (void)
{
  d = f;
  f = i;
}
[hjl@gnu-tgl-2 pr87007]$ make CC=gcc
gcc -O2 -march=skylake -mtune-ctrl=use_vector_fp_converts,use_vector_converts
-S x.c
[hjl@gnu-tgl-2 pr87007]$ cat x.s
        .file   "x.c"
        .text
        .p2align 4
        .globl  foo
        .type   foo, @function
foo:
.LFB0:
        .cfi_startproc
        vxorps  %xmm0, %xmm0, %xmm0
        vcvtss2sd       f(%rip), %xmm0, %xmm1
        vcvtsi2ssl      i(%rip), %xmm0, %xmm0
        vmovsd  %xmm1, d(%rip)
        vmovss  %xmm0, f(%rip)
        ret
        .cfi_endproc
.LFE0:
        .size   foo, .-foo
        .ident  "GCC: (GNU) 11.2.1 20210728 (Red Hat 11.2.1-1)"
        .section        .note.GNU-stack,"",@progbits
[hjl@gnu-tgl-2 pr87007]$

[Bug target/101900] -mtune-ctrl=use_vector_fp_converts,use_vector_converts doesn't work with AVX

[hjl.tools at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101900

--- Comment #1 from H.J. Lu <hjl.tools at gmail dot com> ---
Created attachment 51301
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=51301&action=edit
A patch

[Bug target/101900] -mtune-ctrl=use_vector_fp_converts,use_vector_converts doesn't work with AVX

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101900

--- Comment #2 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:16cca1806d2ac6f03cdc8942edd8762dab1851c4

commit r12-3614-g16cca1806d2ac6f03cdc8942edd8762dab1851c4
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Wed Sep 15 14:17:58 2021 +0800

    x86: Properly handle USE_VECTOR_FP_CONVERTS/USE_VECTOR_CONVERTS

    Check TARGET_USE_VECTOR_FP_CONVERTS or TARGET_USE_VECTOR_CONVERTS when
    handling avx_partial_xmm_update attribute.  Don't convert AVX partial
    XMM register update if vector packed SSE conversion should be used.

    gcc/

            PR target/101900
            * config/i386/i386-features.c (remove_partial_avx_dependency):
            Check TARGET_USE_VECTOR_FP_CONVERTS and TARGET_USE_VECTOR_CONVERTS
            before generating vxorps.

    gcc/testsuite

            PR target/101900
            * gcc.target/i386/pr101900-1.c: New test.
            * gcc.target/i386/pr101900-2.c: Likewise.
            * gcc.target/i386/pr101900-3.c: Likewise.