Re: [Patch] MIPS: mips16e machine patterns - zeb/zeh seb/seh

[Ian Lance Taylor <ian at airs dot com>]

David Ung <davidu@mips.com> writes:

> > I suppose there's the question of whether this should be handled by a
> > separate switch (-mips16e, like you've done) or whether the availability
> > of MIPS16e should be inferred from the combination of -mips16 and the
> > -march setting.
> 
> > I've no strong opinion either way, but I'd like to make sure that the
> > possibility has been considered.
> 
> yes, this is a good question.  Well, mips16e is the 16bit ASE required
> for mips32/64.  So if compiling for mips32/64 and -mips16, it would
> imply -mips16e.  I guess the -mips16e option is useful for non mips32/64
> targets that decides to implement the full mips16e set, though we could
> do without it.  Its probably good to have it because it'll be a similar
> option in gas.  Anyone else have any opionions on this?

My opinion is that, given how MIPS manufacturers seem to mix and match
instruction sets, having the -mips16e option makes sense.  I think
that -march=mips32 -mips16 should imply -mips16e (we can then easily
add -mno-mips16e later when it becomes necessary).

Ian