https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91114
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
--- Comment #4 from Richard Biener <rguenth at gcc dot gnu.org> ---
So the assert says that 'fatal' issues should pop up with _all_ vector sizes.
This is to avoid missed optimizations where the first tried one says 'fatal'
but a later tried one would not (so it's not a correctness assert).
So the first run with vector size 16 doesn't fail fatally but the second with
vector_size 8 does because there isn't a vector type to vectorize 'double'
(so we don't consider V1DF in mode_for_vector (inner_mode, nunits)).
So having a vector type for some stmt isn't really a fatal error.
