https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99037
Bug ID: 99037
Summary: Invalid representation of vector zero in
aarch64-simd.md
Product: gcc
Version: 11.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: rsandifo at gcc dot gnu.org
CC: andre.simoesdiasvieira at arm dot com
Target Milestone: ---
Target: aarch64*-*-*
As Andre noticed, the following patterns have an invalid
representation of constant zero:
- move_lo_quad_internal_<mode>
- move_lo_quad_internal_be_<mode>
The vector forms use (vec_duplicate …) of a constant instead
of a const_vector, and both forms use (const_int 0) even for
FP modes.
In r5-3343 the patterns were split into two, one with a
vec_duplicate and one with a const_int. But I think that
was the wrong (partial) fix: we should have a single pattern
for each, but use aarch64_simd_or_scalar_imm_zero to match
the right form.
I guess this will have to wait for GCC 12 at this point.
