On Wed, 6 Apr 2011, Paul Koning wrote:
> On Apr 6, 2011, at 4:56 PM, Joseph S. Myers wrote:
>
> > On Wed, 6 Apr 2011, Paul Koning wrote:
> >
> >> In i386-gcc 4.5.1, --target-help says that "generic32" and "generic64"
> >> are valid values for -march= and -mtune. In fact, those values are
> >> rejected (even though there seems to be code in i386.c to handle those
> >> values).
> >
> > See PR 45731. This is help for assembler options, not compiler options.
>
> But mtune= is a compiler switch, so why would --target-help describe
> "generic64" as a valid choice for that switch when it is not?
Because it is a valid choice for the assembler switch of the same name as
the compiler switch. Look at the correct output from a trunk compiler:
The following options are target specific:
[...]
-mtune= Schedule code for given CPU
[...]
Assembler options
=================
Use "-Wa,OPTION" to pass "OPTION" to the assembler.
[...]
-mtune=CPU optimize for CPU, CPU is one of:
i8086, i186, i286, i386, i486, pentium, pentiumpro,
pentiumii, pentiumiii, pentium4, prescott, nocona,
core, core2, k6, k6_2, athlon, k8, amdfam10,
generic32, generic64
[...]
Using -Wa,-mtune=generic64 works fine.
--
Joseph S. Myers
jos...@codesourcery.com
