[Bug target/85918] Conversions to/from [unsigned] long long are not vectorized for AVX512DQ target

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85918

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2018-05-25
           Assignee|unassigned at gcc dot gnu.org      |jakub at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Created attachment 44188
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=44188&action=edit
gcc9-pr85918.patch

Partial fix.  This just renames some patterns so that they are matched by the
optab, which is looking for floatuns<mode>2 or fixuns_trunc<mode>2, rather than
ufloat<mode>2 or ufix_trunc<mode>2 which the backend was using.  This fixes
ull2d and d2ull vectorization.
I'll look at the rest later.