Irix6 long doubles implemented wrong? (27_io/ostream_inserter_arith)

[Alexandre Oliva]

On Jan  7, 2003, Richard Henderson <rth@redhat.com> wrote:

>> * optabs.c (expand_binop) <add, sub>: Return xtarget if we haven't
>> been able to move the result to target.

> Ok, I guess.  How does this come up?

There's no movti insn, so the test that decides whether it's possible
to move the result to the final location fails and we omit the move at
that point.  It ends up being emitted as a multi-insn move sequence
upstream, as long as we return the correct location of the result.

FWIW, I had movti patterns in my patch at some point, that split into
pairs of DI moves, but it proved to be pointless, so I took it out.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer