r278235 broke conversions of vector/scalar shifts into vector/vector shifts on targets that only provide the latter. We need to record whether a conversion is required in that case too.
Also, the old useless_type_conversion_p condition seemed unnecessarily strong, since the shift amount can have a different signedness from the shifted value and its vector type is never assumed to be identical to vectype. The patch therefore uses tree_nop_conversion_p instead. Tested on aarch64-linux-gnu and spot-checked on powerpc64-linux-gnu. OK to install? Richard 2019-11-15 Richard Sandiford <[email protected]> gcc/ PR tree-optimization/92515 * tree-vect-stmts.c (vectorizable_shift): Record incompatible op1 types when converting a vector/scalar shift into a vector/vector one, using tree_nop_conversion_p instead of useless_type_conversion_p. Move the conversion code to the transform block. Index: gcc/tree-vect-stmts.c =================================================================== --- gcc/tree-vect-stmts.c 2019-11-14 15:31:21.616704454 +0000 +++ gcc/tree-vect-stmts.c 2019-11-15 12:54:19.281608678 +0000 @@ -5750,21 +5750,9 @@ vectorizable_shift (stmt_vec_info stmt_i the rhs being int, instead of the same type as the lhs, so make sure the scalar is the right type if we are dealing with vectors of long long/long/short/char. */ - if (dt[1] == vect_constant_def) - { - if (!slp_node) - op1 = fold_convert (TREE_TYPE (vectype), op1); - } - else if (!useless_type_conversion_p (TREE_TYPE (vectype), - TREE_TYPE (op1))) - { - if (vec_stmt && !slp_node) - { - op1 = fold_convert (TREE_TYPE (vectype), op1); - op1 = vect_init_vector (stmt_info, op1, - TREE_TYPE (vectype), NULL); - } - } + incompatible_op1_vectype_p + = !tree_nop_conversion_p (TREE_TYPE (vectype), + TREE_TYPE (op1)); } } } @@ -5819,6 +5807,14 @@ vectorizable_shift (stmt_vec_info stmt_i dump_printf_loc (MSG_NOTE, vect_location, "transform binary/unary operation.\n"); + if (incompatible_op1_vectype_p && !slp_node) + { + op1 = fold_convert (TREE_TYPE (vectype), op1); + if (dt[1] != vect_constant_def) + op1 = vect_init_vector (stmt_info, op1, + TREE_TYPE (vectype), NULL); + } + /* Handle def. */ vec_dest = vect_create_destination_var (scalar_dest, vectype);
