https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123175
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
Status|NEW |ASSIGNED
Priority|P3 |P2
--- Comment #16 from Richard Biener <rguenth at gcc dot gnu.org> ---
No, it's not fixed besides for the two actual testcases. I had more places
"fixed" w/o testcases but given only aarch64 can ever create those VEC_PERMs
right now(?) without the help of the GIMPLE FE I felt it was less important
at this point. Also the x86 backend cannot code-gen any of the problematic
VEC_PERMs, so it would get lowered to scalars (lowering to zero-extended
version is missing for more general use).
All vec_perm_indices CTOR uses have to be audited (or as I proposed, make
an overload from a gassign - but that felt like stage1 material). Let me
take this bug again, for this part.
