[Bug driver/60233] New: AVX instructions emitted with -march=native on host without AVX support

[Bug driver/60233] New: AVX instructions emitted with -march=native on host without AVX support

[steven at uplinklabs dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

            Bug ID: 60233
           Summary: AVX instructions emitted with -march=native on host
                    without AVX support
           Product: gcc
           Version: 4.8.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: driver
          Assignee: unassigned at gcc dot gnu.org
          Reporter: steven at uplinklabs dot net
              Host: x86_64-unknown-linux-gnu
            Target: x86_64-unknown-linux-gnu
             Build: x86_64-unknown-linux-gnu

Running GCC in a paravirtual Xen guest on an Ivy Bridge host. The PV guest does
not have AVX support, and -march=native seems to detect this correctly:

$ gcc -O2 -march=native -E -v - </dev/null 2>&1 | grep cc1
 /usr/lib/gcc/x86_64-unknown-linux-gnu/4.8.2/cc1 -E -quiet -v -
-march=core-avx-i -mcx16 -msahf -mno-movbe -maes -mpclmul -mpopcnt -mno-abm
-mno-lwp -mno-fma -mno-fma4 -mno-xop -mno-bmi -mno-bmi2 -mno-tbm -mno-avx
-mno-avx2 -msse4.2 -msse4.1 -mno-lzcnt -mno-rtm -mno-hle -mrdrnd -mf16c
-mfsgsbase -mno-rdseed -mno-prfchw -mno-adx -mfxsr -mno-xsave -mno-xsaveopt
--param l1-cache-size=32 --param l1-cache-line-size=64 --param
l2-cache-size=25600 -mtune=core-avx-i -O2


Note in particular the -mno-avx and -mno-xsave.

However, GCC is still emitting AVX instructions:

$ cat vadd8.c
typedef float v8f __attribute__ ((vector_size(32), aligned(1)));

void vadd8f(v8f *o, v8f *a, v8f *b)
{
        *o = *a + *b;
}

$ echo | gcc -dM -E - -march=native | grep AVX
#define __AVX__ 1

$ gcc -O2 -march=native -c vadd8.c

$ objdump -S vadd8.o

vadd8.o:     file format elf64-x86-64


Disassembly of section .text:

0000000000000000 <vadd8f>:
   0:   c5 f8 10 0e             vmovups (%rsi),%xmm1
   4:   c5 f8 10 02             vmovups (%rdx),%xmm0
   8:   c4 e3 75 18 4e 10 01    vinsertf128 $0x1,0x10(%rsi),%ymm1,%ymm1
   f:   c4 e3 7d 18 42 10 01    vinsertf128 $0x1,0x10(%rdx),%ymm0,%ymm0
  16:   c5 f4 58 c0             vaddps %ymm0,%ymm1,%ymm0
  1a:   c5 f8 11 07             vmovups %xmm0,(%rdi)
  1e:   c4 e3 7d 19 47 10 01    vextractf128 $0x1,%ymm0,0x10(%rdi)
  25:   c5 f8 77                vzeroupper 
  28:   c3                      retq   

I'm guessing that the -march=core-avx-i it decided on is overriding the
-mno-avx flag, but this seems to be a regression to me.

Version information:

$ gcc -v
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-unknown-linux-gnu/4.8.2/lto-wrapper
Target: x86_64-unknown-linux-gnu
Configured with: /build/gcc/src/gcc-4.8-20140206/configure --prefix=/usr
--libdir=/usr/lib --libexecdir=/usr/lib --mandir=/usr/share/man
--infodir=/usr/share/info --with-bugurl=https://bugs.archlinux.org/
--enable-languages=c,c++,ada,fortran,go,lto,objc,obj-c++ --enable-shared
--enable-threads=posix --with-system-zlib --enable-__cxa_atexit
--disable-libunwind-exceptions --enable-clocale=gnu --disable-libstdcxx-pch
--disable-libssp --enable-gnu-unique-object --enable-linker-build-id
--enable-cloog-backend=isl --disable-cloog-version-check --enable-lto
--enable-plugin --enable-install-libiberty --with-linker-hash-style=gnu
--disable-multilib --disable-werror --enable-checking=release
Thread model: posix
gcc version 4.8.2 20140206 (prerelease) (GCC)

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Can you attach /pro/cpuinfo?

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[steven at uplinklabs dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #2 from Steven Noonan <steven at uplinklabs dot net> ---
Created attachment 32147
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32147&action=edit
/proc/cpuinfo

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[steven at uplinklabs dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #3 from Steven Noonan <steven at uplinklabs dot net> ---
Created attachment 32148
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32148&action=edit
CPUID dump

Adding both /proc/cpuinfo and a dump from a userspace utility using the CPUID
instruction (since /proc/cpuinfo is assembled with information obtained via PV
CPUID hypercall).

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-02-17
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |uros at gcc dot gnu.org
   Target Milestone|---                         |4.8.3
     Ever confirmed|0                           |1

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I think the reason for this is that -march=native passes in your case -mf16c,
and -mf16c implies -mavx.  So, either OPTION_MASK_ISA_F16C_SET should not
include OPTION_MASK_ISA_AVX_SET, or the driver shouldn't set -mf16c if AVX
support is missing.
As at least some of the F16C instructions use ymmN registers, if we'd change
OPTION_MASK_ISA_F16C_SET, then the *256 TARGET_F16C patterns would also need to
be guarded with && TARGET_AVX.
For the latter alternative, we would need to do something like:
--- gcc/config/i386/driver-i386.c 2014-01-03 11:41:06.393269411 +0100
+++ gcc/config/i386/driver-i386.c 2014-02-17 07:32:41.289022308 +0100
@@ -513,6 +513,7 @@ const char *host_detect_local_cpu (int a
       has_avx2 = 0;
       has_fma = 0;
       has_fma4 = 0;
+      has_f16c = 0;
       has_xop = 0;
       has_xsave = 0;
       has_xsaveopt = 0;

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #5 from Uroš Bizjak <ubizjak at gmail dot com> ---
(In reply to Jakub Jelinek from comment #4)
> I think the reason for this is that -march=native passes in your case
> -mf16c, and -mf16c implies -mavx.  So, either OPTION_MASK_ISA_F16C_SET
> should not include OPTION_MASK_ISA_AVX_SET, or the driver shouldn't set
> -mf16c if AVX support is missing.
> As at least some of the F16C instructions use ymmN registers, if we'd change
> OPTION_MASK_ISA_F16C_SET, then the *256 TARGET_F16C patterns would also need
> to be guarded with && TARGET_AVX.
> For the latter alternative, we would need to do something like:
> --- gcc/config/i386/driver-i386.c 2014-01-03 11:41:06.393269411 +0100
> +++ gcc/config/i386/driver-i386.c 2014-02-17 07:32:41.289022308 +0100
> @@ -513,6 +513,7 @@ const char *host_detect_local_cpu (int a
>        has_avx2 = 0;
>        has_fma = 0;
>        has_fma4 = 0;
> +      has_f16c = 0;
>        has_xop = 0;
>        has_xsave = 0;
>        has_xsaveopt = 0;

This is the correct approach. We already disable f16c for -mno-avx in
common/config/i386/i386-common.c in this way, and it looks that driver-i386.c
was not updated accordingly.

There are no real processors with F16C and no AVX, but we should be consistent
here and follow i386-common.c.
>From gcc-bugs-return-443877-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Mon Feb 17 08:52:24 2014
Return-Path: <gcc-bugs-return-443877-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 1509 invoked by alias); 17 Feb 2014 08:52:23 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 1472 invoked by uid 48); 17 Feb 2014 08:52:19 -0000
From: "ubizjak at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support
Date: Mon, 17 Feb 2014 08:52:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: driver
X-Bugzilla-Version: 4.8.2
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: ubizjak at gmail dot com
X-Bugzilla-Status: NEW
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: 4.8.3
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-60233-4-V4fEgDGK6q@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-60233-4@http.gcc.gnu.org/bugzilla/>
References: <bug-60233-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2014-02/txt/msg01634.txt.bz2
Content-length: 325

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #6 from Uroš Bizjak <ubizjak at gmail dot com> ---
And while looking at driver-i386.c, it looks to me that the whole osxsave state
check should be moved below (ext_level > 0x80000000) processing, otherwise we
won't clear FMA4 and XOP flags correctly.
>From gcc-bugs-return-443878-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org Mon Feb 17 08:58:40 2014
Return-Path: <gcc-bugs-return-443878-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Delivered-To: listarch-gcc-bugs@gcc.gnu.org
Received: (qmail 4822 invoked by alias); 17 Feb 2014 08:58:39 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Delivered-To: mailing list gcc-bugs@gcc.gnu.org
Received: (qmail 4785 invoked by uid 48); 17 Feb 2014 08:58:34 -0000
From: "mikpelinux at gmail dot com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug tree-optimization/60229] wrong code at -O2 and -O3 on x86_64-linux-gnu
Date: Mon, 17 Feb 2014 08:58:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: tree-optimization
X-Bugzilla-Version: 4.7.4
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: mikpelinux at gmail dot com
X-Bugzilla-Status: RESOLVED
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: unassigned at gcc dot gnu.org
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-60229-4-gVnjs4har9@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-60229-4@http.gcc.gnu.org/bugzilla/>
References: <bug-60229-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2014-02/txt/msg01635.txt.bz2
Content-length: 361

http://gcc.gnu.org/bugzilla/show_bug.cgi?id`229

--- Comment #3 from Mikael Pettersson <mikpelinux at gmail dot com> ---
Technically there is an overflow there.  But GCC defines conversion to a
smaller signed integer type, when the value cannot be represented in that
smaller type, as a non-signalling truncation.  Still, portable code mustn't
rely on that.

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 32151
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32151&action=edit
gcc49-pr60233.patch

Untested fix.

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Tue Feb 18 07:19:46 2014
New Revision: 207833

URL: http://gcc.gnu.org/viewcvs?rev=207833&root=gcc&view=rev
Log:
    PR driver/60233
    * config/i386/driver-i386.c (host_detect_local_cpu): If
    YMM state is not saved by the OS, also clear has_f16c.  Move
    CPUID 0x80000001 handling before YMM state saving checking.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/i386/driver-i386.c

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #9 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Tue Feb 18 07:23:51 2014
New Revision: 207834

URL: http://gcc.gnu.org/viewcvs?rev=207834&root=gcc&view=rev
Log:
    PR driver/60233
    * config/i386/driver-i386.c (host_detect_local_cpu): If
    YMM state is not saved by the OS, also clear has_f16c.  Move
    CPUID 0x80000001 handling before YMM state saving checking.

Modified:
    branches/gcc-4_8-branch/gcc/ChangeLog
    branches/gcc-4_8-branch/gcc/config/i386/driver-i386.c

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #11 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Author: jakub
Date: Tue Feb 18 07:32:17 2014
New Revision: 207835

URL: http://gcc.gnu.org/viewcvs?rev=207835&root=gcc&view=rev
Log:
    PR driver/60233
    * config/i386/driver-i386.c (host_detect_local_cpu): If
    YMM state is not saved by the OS, also clear has_f16c.  Move
    CPUID 0x80000001 handling before YMM state saving checking.

Modified:
    branches/gcc-4_7-branch/gcc/ChangeLog
    branches/gcc-4_7-branch/gcc/config/i386/driver-i386.c

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

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

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

--- Comment #10 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Should be fixed now.

[Bug driver/60233] AVX instructions emitted with -march=native on host without AVX support

[steven at uplinklabs dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60233

--- Comment #12 from Steven Noonan <steven at uplinklabs dot net> ---
Thanks for the fast resolution, Jakub!