https://gcc.gnu.org/bugzilla/show_bug.cgi?id=105816
Bug ID: 105816
Summary: SLP vectorize permutes from different groups
Product: gcc
Version: 13.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: rguenth at gcc dot gnu.org
Target Milestone: ---
typedef int v4si __attribute__((vector_size(16)));
typedef int v8si __attribute__((vector_size(32)));
void __attribute__((noipa)) test_lo (v8si *dst, v4si src1, v4si src2)
{
(*dst)[0] = src1[0];
(*dst)[1] = src1[1];
(*dst)[2] = src1[2];
(*dst)[3] = src1[3];
(*dst)[4] = src2[0];
(*dst)[5] = src2[1];
(*dst)[6] = src2[2];
(*dst)[7] = src2[3];
}
should now be vectorizable but we reject this because we are not able
to split the groups only at the point where it is necessary, merging
them with a permute node. Instead we fail SLP discovery and rely on
the toplevel logic of splitting the group at the stores.
