On Tue, Jul 24, 2018 at 11:54 AM Richard Sandiford <richard.sandif...@arm.com> wrote: > > The usual vectoriser dance to create new assignments is: > > new_stmt = gimple_build_assign (vec_dest, ...); > new_temp = make_ssa_name (vec_dest, new_stmt); > gimple_assign_set_lhs (new_stmt, new_temp); > > but one site in vectorizable_reduction used: > > new_temp = make_ssa_name (vec_dest, new_stmt); > > before creating new_stmt. > > This method of creating statements probably needs cleaning up, but > that's for another day...
Yeah, one can elide the set_lhs by first allocating the SSA name without defining stmt and then building the stmt with the SSA def directly... OK > > 2018-07-24 Richard Sandiford <richard.sandif...@arm.com> > > gcc/ > * tree-vect-loop.c (vectorizable_reduction): Fix an instance in > which make_ssa_name was called with new_stmt before new_stmt > had been created. > > Index: gcc/tree-vect-loop.c > =================================================================== > --- gcc/tree-vect-loop.c 2018-07-24 10:22:12.737465897 +0100 > +++ gcc/tree-vect-loop.c 2018-07-24 10:22:16.421433184 +0100 > @@ -7210,9 +7210,10 @@ vectorizable_reduction (gimple *stmt, gi > if (op_type == ternary_op) > vop[2] = vec_oprnds2[i]; > > - new_temp = make_ssa_name (vec_dest, new_stmt); > - new_stmt = gimple_build_assign (new_temp, code, > + new_stmt = gimple_build_assign (vec_dest, code, > vop[0], vop[1], vop[2]); > + new_temp = make_ssa_name (vec_dest, new_stmt); > + gimple_assign_set_lhs (new_stmt, new_temp); > } > vect_finish_stmt_generation (stmt, new_stmt, gsi); >
