https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125197
Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|WAITING |NEW
--- Comment #15 from Eric Botcazou <ebotcazou at gcc dot gnu.org> ---
> For x86_64-linux, doesn't the -m32 flag imply -march=x86-64 i.e. the 32-bit
> instructions supported by all x86_64 CPUs? If you want to use an x86_64
> multilib compiler to produce code for i486 then you need to say so
> explicitly, the -m32 option does not default to producing code that will run
> on non-x86_64 CPUs.
Are you sure that this was the case historically on x86-64? AFAICS the
original logic in configure.tgt for x86-64 -m32 was to force -march=i486,
exactly like -mcpu=v9 is forced for SPARC64 (this was only tweaked in 2021 by
Jakub).
> Is the sparcv9 -m32 multilib different? Does it really need to (*by
> default*) produce code that will run on pre-sparcv9 CPUs?
Historically, yes, like x86-64 I think. The problem with changing it now is
that the testing coverage will be very low.
