[Bug target/57333] Wrong detection of LZCNT instruction, -mno-lzcnt has no effect

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57333

Andrew Pinski  changed:

   What|Removed |Added

 Ever confirmed|0   |1
 Status|UNCONFIRMED |WAITING

--- Comment #6 from Andrew Pinski  ---
(In reply to Andrew Pinski from comment #5)
> Testcase for the error:
> $ echo | gcc -dM -E - -march=amdfam10 -mno-lzcnt | grep LZCNT
> #define __LZCNT__ 1
> 
> 
> Is broken again on the trunk.

Filed PR 101685 for that.


As far as the LTO issue, it might be already fixed a different way; of course
without a testcase it is hard to say why it is broken.

[Bug target/57333] Wrong detection of LZCNT instruction, -mno-lzcnt has no effect

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57333

Andrew Pinski  changed:

   What|Removed |Added

 Ever confirmed|1   |0
 Status|NEW |UNCONFIRMED

[Bug target/57333] Wrong detection of LZCNT instruction, -mno-lzcnt has no effect

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57333

Andrew Pinski  changed:

   What|Removed |Added

 Target||x86_64-linux-gnu
   Last reconfirmed||2021-07-30
 Ever confirmed|0   |1
 Status|UNCONFIRMED |NEW

--- Comment #5 from Andrew Pinski  ---
Testcase for the error:
$ echo | gcc -dM -E - -march=amdfam10 -mno-lzcnt | grep LZCNT
#define __LZCNT__ 1


Is broken again on the trunk.

[Bug target/57333] Wrong detection of LZCNT instruction, -mno-lzcnt has no effect

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57333

Richard Biener rguenth at gcc dot gnu.org changed:

   What|Removed |Added

   Target Milestone|4.7.4   |---

--- Comment #4 from Richard Biener rguenth at gcc dot gnu.org ---
Unsetting target milestone of open non-regression bug from version of branch
being closed.

[Bug target/57333] Wrong detection of LZCNT instruction, -mno-lzcnt has no effect

[ubizjak at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57333

Uroš Bizjak ubizjak at gmail dot com changed:

   What|Removed |Added

 CC||tmsriram at google dot com
   Target Milestone|--- |4.7.4
   Severity|major   |normal

--- Comment #2 from Uroš Bizjak ubizjak at gmail dot com ---
LZCNT insn is part of ABM abi, and is enabled with -mabm. Does LZCNT insn
really cause SIGILL on your target?

The issue with -mno-lzcnt will be fixed as part of patch at [1].

[1] http://gcc.gnu.org/ml/gcc-patches/2013-05/msg01054.html

[Bug target/57333] Wrong detection of LZCNT instruction, -mno-lzcnt has no effect

[bernhard.hartleb at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57333

--- Comment #3 from bernhard.hartleb at gmail dot com ---
(In reply to Uroš Bizjak from comment #2)
 LZCNT insn is part of ABM abi, and is enabled with -mabm. Does LZCNT insn
 really cause SIGILL on your target?

No sorry, this was an error on my part, because I did not know about -mabm
before.

The compilation of (libQtGui.so.4.8.4) with -march=k8 finished only after
reporting this bug and it still produced the illegal instruction. There is no
problem with LZCNT, other than -mno-lzcnt not working.

My issue is somewhere with LTO, since the build works percectly with LTO
disabled.
But this will be a different bug report.

[Bug target/57333] Wrong detection of LZCNT instruction, -mno-lzcnt has no effect

[bernhard.hartleb at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57333

--- Comment #1 from bernhard.hartleb at gmail dot com ---
It seems the illegal instructions is caused by the use of LTO.
Without LTO everything is fine.

The second point is still valid, however.