------- Additional Comments From tkoenig at gcc dot gnu dot org 2005-04-23 18:26 ------- (In reply to comment #2) > Or, alternatively, we could call reshape_generic. Which of course would be > evil > from a performance standpoitn :-)
Actually, that's what we are doing for a lot of intrinsics: cshift1.m4: memcpy (dest, src, size); eoshift1.m4: memcpy (dest, src, size); eoshift1.m4: memcpy (dest, pbound, size); eoshift3.m4: memcpy (dest, src, size); eoshift3.m4: memcpy (dest, bptr, size); in_unpack.m4: memcpy (dest, src, dsize * rtype_kind); apparently to fix that same bug. For better performance, these should be fixed as well. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21127