[Bug c/104673] New: powerpc e500mc Error: unrecognized opcode: `isel'

[Bug c/104673] New: powerpc e500mc Error: unrecognized opcode: `isel'

[judge.packham at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104673

            Bug ID: 104673
           Summary: powerpc e500mc Error: unrecognized opcode: `isel'
           Product: gcc
           Version: 11.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: judge.packham at gmail dot com
  Target Milestone: ---

Created attachment 52503
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52503&action=edit
assembly file emitted by GCC

When building for powerpc with -mcpu=e500mc the generated assembly includes a
".machine ppc" directive which causes gas to ignore the command line -me500mc
flag. This means that it decides any e500mc specific instructions are invalid.

Attached is the generated assembly from a build with gcc 11.2 cross-compiled
for powerpc. As you can see the .machine directive says ppc but when gas sees
this it will reset it's internal state and decide the isel instruction is
invalid.

I think this might be an regression from
https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=2d94f7dea9.

GCC only seems to have a limited set of machines it will emit which don't match
with what binutils uses.

[Bug target/104673] powerpc e500mc Error: unrecognized opcode: `isel'

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104673

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
How did you configure GCC? How did you invoke GCC?
Can you attach the preprocessed source rather than the generated assembly code?

[Bug target/104673] powerpc e500mc Error: unrecognized opcode: `isel'

[judge.packham at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104673

--- Comment #2 from Chris Packham <judge.packham at gmail dot com> ---
GCC configure line is

$ /home/ctng/crosstool-ng/.build/powerpc-unknown-linux-gnu/src/gcc/configure
--build=x86_64-build_pc-linux-gnu --host=x86_64-build_pc-linux-gnu
--target=powerpc-unknown-linux-gnu
--prefix=/home/ctng/x-tools/powerpc-unknown-linux-gnu
--exec_prefix=/home/ctng/x-tools/powerpc-unknown-linux-gnu
--with-sysroot=/home/ctng/x-tools/powerpc-unknown-linux-gnu/powerpc-unknown-linux-gnu/sysroot
--enable-languages=c,c++ --with-cpu=e500mc --with-pkgversion=crosstool-NG
1.24.0.548_0e1989a --enable-__cxa_atexit --disable-libmudflap --disable-libgomp
--disable-libssp --disable-libquadmath --disable-libquadmath-support
--disable-libsanitizer --disable-libmpx --disable-libstdcxx-verbose
--with-gmp=/home/ctng/crosstool-ng/.build/powerpc-unknown-linux-gnu/buildtools
--with-mpfr=/home/ctng/crosstool-ng/.build/powerpc-unknown-linux-gnu/buildtools
--with-mpc=/home/ctng/crosstool-ng/.build/powerpc-unknown-linux-gnu/buildtools
--with-isl=/home/ctng/crosstool-ng/.build/powerpc-unknown-linux-gnu/buildtools
--enable-lto --enable-threads=posix --enable-target-optspace --enable-plugin
--disable-nls --disable-multilib
--with-local-prefix=/home/ctng/x-tools/powerpc-unknown-linux-gnu/powerpc-unknown-linux-gnu/sysroot
--enable-long-long

Invocation is

$ powerpc-unknown-linux-gnu-gcc -mcpu=e500mc -O2 -c testcase.c -S -o testcase.S

The original error occurred when building glibc. I've yet to produce a testcase
that triggers gcc to emit the isel instruction but you can see the .machine
directive doesn't match up with -mcpu or what gets passed to gas.

[Bug target/104673] powerpc e500mc Error: unrecognized opcode: `isel'

[judge.packham at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104673

--- Comment #3 from Chris Packham <judge.packham at gmail dot com> ---
Created attachment 52511
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=52511&action=edit
Attempt to trigger problem

[Bug target/104673] powerpc e500mc Error: unrecognized opcode: `isel'

[judge.packham at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104673

Chris Packham <judge.packham at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #4 from Chris Packham <judge.packham at gmail dot com> ---
Looks like a duplicate of https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104090

*** This bug has been marked as a duplicate of bug 104090 ***