[Bug middle-end/68134] [6 Regression] float64x1_t comparison ICE on aarch64-none-elf

[ienkovich at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68134

Ilya Enkovich <ienkovich at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |ienkovich at gcc dot gnu.org

--- Comment #4 from Ilya Enkovich <ienkovich at gcc dot gnu.org> ---
Seems the problem is that we have V1DF mode but don't have V1DI mode.  It
causes mode_for_vector to return DI instead of V1DI which makes vector lowering
pass think it is a scalar statements which doesn't need lowering.  This patch
should help:

diff --git a/gcc/targhooks.c b/gcc/targhooks.c
index c34b4e9..66d983b 100644
--- a/gcc/targhooks.c
+++ b/gcc/targhooks.c
@@ -1093,8 +1093,8 @@ default_get_mask_mode (unsigned nunits, unsigned
vector_size)
   gcc_assert (elem_size * nunits == vector_size);

   vector_mode = mode_for_vector (elem_mode, nunits);
-  if (VECTOR_MODE_P (vector_mode)
-      && !targetm.vector_mode_supported_p (vector_mode))
+  if (!VECTOR_MODE_P (vector_mode)
+      || !targetm.vector_mode_supported_p (vector_mode))
     vector_mode = BLKmode;

   return vector_mode;