------- Comment #2 from dave at hiauly1 dot hia dot nrc dot ca 2010-05-24 18:11 ------- Subject: Re: Multiplying -1 by NaN is not valid.
On Mon, 24 May 2010, rguenth at gcc dot gnu dot org wrote: > Well - GCC has fallback expansions for some sign-related instructions by doing > bit-fiddling instead. I think that's whats required if the arch cannot do > a IEEE negate. I am testing the attached change to fiddle the sign bit. It is probably more efficient than a generic fallback. Dave ------- Comment #3 from dave at hiauly1 dot hia dot nrc dot ca 2010-05-24 18:11 ------- Created an attachment (id=20737) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=20737&action=view) -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44261