[Bug math/28252] New: Unnecessary memory load in AVX512 SVML codes

[Bug math/28252] New: Unnecessary memory load in AVX512 SVML codes

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28252

            Bug ID: 28252
           Summary: Unnecessary memory load in AVX512 SVML codes
           Product: glibc
           Version: 2.35
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
          Assignee: haochen.jiang at intel dot com
          Reporter: hjl.tools at gmail dot com
  Target Milestone: ---

AVX512 SVML codes have unnecessary memory load. For example,
svml_s_sincosf16_core_avx512.S has

       vmovups   .L_2il0floatpacket.13(%rip), %zmm14
...
        .section .rodata, "a"
.L_2il0floatpacket.13:
        .long  
0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff,0xffffffff
        .type   .L_2il0floatpacket.13,@object

They can be replaced by

          vpternlogd    $0xFF, %zmm0, %zmm0, %zmm0

Any particular reason to use memory load?

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/28252] Unnecessary memory load in AVX512 SVML codes

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28252

--- Comment #1 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by H.J. Lu <hjl@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=78c9ec9000f873abe7a15a91b87080a2e4308260

commit 78c9ec9000f873abe7a15a91b87080a2e4308260
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Fri Aug 20 06:42:24 2021 -0700

    x86-64: Optimize load of all bits set into ZMM register [BZ #28252]

    Optimize loads of all bits set into ZMM register in AVX512 SVML codes
    by replacing

            vpbroadcastq .L_2il0floatpacket.16(%rip), %zmmX

    and

            vmovups   .L_2il0floatpacket.13(%rip), %zmmX

    with
            vpternlogd $0xff, %zmmX, %zmmX, %zmmX

    This fixes BZ #28252.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/28252] Unnecessary memory load in AVX512 SVML codes

[hjl.tools at gmail dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=28252

H.J. Lu <hjl.tools at gmail dot com> changed:

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

--- Comment #2 from H.J. Lu <hjl.tools at gmail dot com> ---
Fixed for 2.35.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/28252] Unnecessary memory load in AVX512 SVML codes

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28252

--- Comment #3 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
The release/2.34/master branch has been updated by Sunil Pandey
<skpgkp2@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=16245986fb9bfe396113fc7dfd1929f69a9e748e

commit 16245986fb9bfe396113fc7dfd1929f69a9e748e
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Fri Aug 20 06:42:24 2021 -0700

    x86-64: Optimize load of all bits set into ZMM register [BZ #28252]

    Optimize loads of all bits set into ZMM register in AVX512 SVML codes
    by replacing

            vpbroadcastq .L_2il0floatpacket.16(%rip), %zmmX

    and

            vmovups   .L_2il0floatpacket.13(%rip), %zmmX

    with
            vpternlogd $0xff, %zmmX, %zmmX, %zmmX

    This fixes BZ #28252.

    (cherry picked from commit 78c9ec9000f873abe7a15a91b87080a2e4308260)

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug math/28252] Unnecessary memory load in AVX512 SVML codes

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28252

--- Comment #4 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
The release/2.33/master branch has been updated by Sunil Pandey
<skpgkp2@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=c09d92d4f657560bf884c735bc8fdd53e4276367

commit c09d92d4f657560bf884c735bc8fdd53e4276367
Author: H.J. Lu <hjl.tools@gmail.com>
Date:   Fri Aug 20 06:42:24 2021 -0700

    x86-64: Optimize load of all bits set into ZMM register [BZ #28252]

    Optimize loads of all bits set into ZMM register in AVX512 SVML codes
    by replacing

            vpbroadcastq .L_2il0floatpacket.16(%rip), %zmmX

    and

            vmovups   .L_2il0floatpacket.13(%rip), %zmmX

    with
            vpternlogd $0xff, %zmmX, %zmmX, %zmmX

    This fixes BZ #28252.

    (cherry picked from commit 78c9ec9000f873abe7a15a91b87080a2e4308260)

-- 
You are receiving this mail because:
You are on the CC list for the bug.