[Bug target/48743] New: -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[Bug target/48743] New: -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[gcc.gnu.org at stellar dot eclipse.co.uk]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

           Summary: -march=native mis-detects AMD K6-2+ / K6-3 as Athlon -
                    compiled C fails with "illegal instruction"
           Product: gcc
           Version: 4.5.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: gcc.gnu.org@stellar.eclipse.co.uk


Created attachment 24084
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24084
Output of  `echo | gcc -dM -E - -march=native`

When I compile nano on my Linux system using `-march=i586`, `-march=k6-2` or
`-march=k6-3` it works fine.
If I compile it with `-march=native` then nano crashes on startup saying
"illegal instruction".


/proc/cpuinfo seems to show the CPU correctly:

# cat /proc/cpuinfo 
processor       : 0
vendor_id       : AuthenticAMD
cpu family      : 5
model           : 13
model name      : AMD-K6(tm)-III Processor
stepping        : 4
cpu MHz         : 448.140
cache size      : 128 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 1
wp              : yes
flags           : fpu vme de pse tsc msr cx8 pge mmx syscall 3dnowext 3dnow
k6_mtrr
bogomips        : 896.28
clflush size    : 32
cache_alignment : 32
address sizes   : 32 bits physical, 32 bits virtual
power management: ts fid vid

However gcc seems to misdetect it:

# gcc -v -Q --help=target -march=native 2>&1 | grep march
/usr/libexec/gcc/i586-pc-linux-gnu/4.5.2/cc1 -v help-dummy -D_FORTIFY_SOURCE=2
-march=athlon --param l1-cache-size=32 --param l1-cache-line-size=32 --param
l2-cache-size=128 -mtune=athlon -dumpbase help-dummy -auxbase help-dummy
-version -fhelp=target -o /tmp/ccQrke9q.s
 -march=                               athlon
# cc -march=native -E -v - </dev/null 2>&1 | grep cc1
 /usr/libexec/gcc/i586-pc-linux-gnu/4.5.2/cc1 -E -quiet -v -
-D_FORTIFY_SOURCE=2 -march=athlon --param l1-cache-size=32 --param
l1-cache-line-size=32 --param l2-cache-size=128 -mtune=athlon

Output of  `echo | gcc -dM -E - -march=native` attached.

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[gcc.gnu.org at stellar dot eclipse.co.uk]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #1 from Joe Stroller <gcc.gnu.org at stellar dot eclipse.co.uk> 2011-04-23 21:12:35 UTC ---
Created attachment 24085
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24085
Output from Todd Allen's cpuid utility

Todd Allen's cpuid utility detects this CPU correctly.
http://www.etallen.com/cpuid.html

Output attached.

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[gcc.gnu.org at stellar dot eclipse.co.uk]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #2 from Joe Stroller <gcc.gnu.org at stellar dot eclipse.co.uk> 2011-04-23 21:14:26 UTC ---
Created attachment 24086
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24086
Raw output from Todd Allen's cpuid utility

Raw output also attached.

I think the AMD document "AMD Processor Recognition Application Note" may also
be relevant:
http://support.amd.com/us/Processor_TechDocs/20734.pdf

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[zuxy.meng at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #3 from Zuxy <zuxy.meng at gmail dot com> 2011-05-12 08:27:56 UTC ---
Created attachment 24233
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24233
Proposed patch

For a brief time AMD produced K6-2+ and K6-3+ processors, targeting mobile
market. These processors support Extended 3DNow! but not CMOV or Extended MMX
(a subset of SSE) and are detected as Athlon by GCC so GCC will generate CMOV
for them, leading to #UD.

The patch solves the problem by differentiating K6 and Athlon by Extended MMX
rather than Extended 3DNow!.

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[mikpe at it dot uu.se]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #4 from Mikael Pettersson <mikpe at it dot uu.se> 2011-05-22 14:23:21 UTC ---
Created attachment 24326
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24326
untested fix

The previous patch doesn't look quite right to me.  The fundamental problem IMO
is that the AMD CPU type detection only looks at feature flags without
considering the CPUID family value.  This (untested) patch qualifies the
ambiguous feature test with a family test before assuming CPU == Athlon.

Does this solve your problem?

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[gcc.gnu.org at stellar dot eclipse.co.uk]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #5 from Joe Stroller <gcc.gnu.org at stellar dot eclipse.co.uk> 2011-06-16 15:10:30 UTC ---
(In reply to comment #4)
> Does this solve your problem?

Yes! Thank you, Mikael, and apologies for not testing sooner.

The fix in attachment 24326 works perfectly.

I tested it on gcc 4.5.2, and the "if" that is changed is at line 509 on that
version, but the structure is exactly the same. 

Now:

# echo | gcc -dM -E - -march=native gcc -v -Q --help=target -march=native 2>&1
| grep march
 /usr/libexec/gcc/i586-pc-linux-gnu/4.5.2/cc1 -E -v help-dummy
-D_FORTIFY_SOURCE=2 -march=k6-3 --param l1-cache-size=32 --param
l1-cache-line-size=32 --param l2-cache-size=128 -mtune=k6 -fhelp=target -dM
  -march=                               k6-3
# cc -march=native -E -v - </dev/null 2>&1 | grep cc1
 /usr/libexec/gcc/i586-pc-linux-gnu/4.5.2/cc1 -E -quiet -v -
-D_FORTIFY_SOURCE=2 -march=k6-3 --param l1-cache-size=32 --param
l1-cache-line-size=32 --param l2-cache-size=128 -mtune=k6
#

Additionally I have recompiled my whole system with -march=native, and it is
working perfectly.

I would be very grateful if this were able to be incorporated into gcc.

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[qneill at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #6 from Quentin Neill <qneill at gcc dot gnu.org> 2011-12-19 17:57:42 UTC ---
Author: qneill
Date: Mon Dec 19 17:57:32 2011
New Revision: 182489

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=182489
Log:
2011-12-19  Quentin Neill  <quentin.neill@amd.com>
2011-12-19  Quentin Neill  <quentin.neill@amd.com>

    PR target/48743
    * config/i386/driver-i386.c (host_detect_local_cpu): Also check
    family to distinguish PROCESSOR_ATHLON.


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

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[qneill at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

Quentin Neill <qneill at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |qneill at gcc dot gnu.org

--- Comment #7 from Quentin Neill <qneill at gcc dot gnu.org> 2011-12-19 18:55:49 UTC ---
Waiting to fix in 4.5 and 4.6 per Uros:
    http://gcc.gnu.org/ml/gcc-patches/2011-12/msg00938.html

Meanwhile, fixed in trunk (4.7) after testing on amdfam10 and bdver1:

Author: qneill
Date: Mon Dec 19 17:57:32 2011
New Revision: 182489

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=182489
Log:
2011-12-19  Quentin Neill  <quentin.neill@amd.com>
2011-12-19  Quentin Neill  <quentin.neill@amd.com>

    PR target/48743
    * config/i386/driver-i386.c (host_detect_local_cpu): Also check
    family to distinguish PROCESSOR_ATHLON.

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

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[qneill at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #8 from Quentin Neill <qneill at gcc dot gnu.org> 2012-01-19 20:15:38 UTC ---
Author: qneill
Date: Thu Jan 19 20:15:29 2012
New Revision: 183309

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=183309
Log:
2012-01-19  Quentin Neill  <quentin.neill@amd.com>

    PR target/48743
    * config/i386/driver-i386.c (host_detect_local_cpu): Also check
    family to distinguish PROCESSOR_ATHLON.


Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/config/i386/driver-i386.c

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[qneill at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

--- Comment #9 from Quentin Neill <qneill at gcc dot gnu.org> 2012-02-07 16:17:35 UTC ---
Author: qneill
Date: Tue Feb  7 16:17:26 2012
New Revision: 183973

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=183973
Log:
2012-02-07  Quentin Neill  <quentin.neill@amd.com>

    Backport from mainline:
    2012-01-19  Quentin Neill  <quentin.neill@amd.com>

    PR target/48743
    * config/i386/driver-i386.c (host_detect_local_cpu): Also check
    family to distinguish PROCESSOR_ATHLON.


Modified:
    branches/gcc-4_5-branch/gcc/ChangeLog
    branches/gcc-4_5-branch/gcc/config/i386/driver-i386.c

[Bug target/48743] -march=native mis-detects AMD K6-2+ / K6-3 as Athlon - compiled C fails with "illegal instruction"

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48743

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.5.4

--- Comment #10 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-02-08 10:45:35 UTC ---
Fixed for 4.5.4.