X-Git-Url: http://git.sourceforge.jp/view?a=blobdiff_plain;f=gcc%2Ftree-vect-patterns.c;h=c1ce14c4bbb91b1aacf84d1b1a6fde43ff0fe11a;hb=ffc0b1edcad561e7c8304c3f6cb38ec31502b024;hp=9fac417d0d0ff3a616b2f0a1752d833db427a4dd;hpb=f031fa031697b65f36bd3d9d3ad4c609577a27be;p=pf3gnuchains%2Fgcc-fork.git diff --git a/gcc/tree-vect-patterns.c b/gcc/tree-vect-patterns.c index 9fac417d0d0..c1ce14c4bbb 100644 --- a/gcc/tree-vect-patterns.c +++ b/gcc/tree-vect-patterns.c @@ -1,12 +1,12 @@ /* Analysis Utilities for Loop Vectorization. - Copyright (C) 2006 Free Software Foundation, Inc. + Copyright (C) 2006, 2007 Free Software Foundation, Inc. Contributed by Dorit Nuzman This file is part of GCC. GCC is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free -Software Foundation; either version 2, or (at your option) any later +Software Foundation; either version 3, or (at your option) any later version. GCC is distributed in the hope that it will be useful, but WITHOUT ANY @@ -15,9 +15,8 @@ FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License -along with GCC; see the file COPYING. If not, write to the Free -Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA -02110-1301, USA. */ +along with GCC; see the file COPYING3. If not see +. */ #include "config.h" #include "system.h" @@ -109,10 +108,6 @@ widened_name_p (tree name, tree use_stmt, tree *half_type, tree *def_stmt) if (!vect_is_simple_use (oprnd0, loop_vinfo, &dummy, &dummy, &dt)) return false; - if (dt != vect_invariant_def && dt != vect_constant_def - && dt != vect_loop_def) - return false; - return true; } @@ -153,7 +148,14 @@ widened_name_p (tree name, tree use_stmt, tree *half_type, tree *def_stmt) * Return value: A new stmt that will be used to replace the sequence of stmts that constitute the pattern. In this case it will be: WIDEN_DOT_PRODUCT -*/ + + Note: The dot-prod idiom is a widening reduction pattern that is + vectorized without preserving all the intermediate results. It + produces only N/2 (widened) results (by summing up pairs of + intermediate results) rather than all N results. Therefore, we + cannot allow this pattern when we want to get all the results and in + the correct order (as is the case when this computation is in an + inner-loop nested in an outer-loop that us being vectorized). */ static tree vect_recog_dot_prod_pattern (tree last_stmt, tree *type_in, tree *type_out) @@ -165,6 +167,8 @@ vect_recog_dot_prod_pattern (tree last_stmt, tree *type_in, tree *type_out) tree type, half_type; tree pattern_expr; tree prod_type; + loop_vec_info loop_info = STMT_VINFO_LOOP_VINFO (stmt_vinfo); + struct loop *loop = LOOP_VINFO_LOOP (loop_info); if (TREE_CODE (last_stmt) != GIMPLE_MODIFY_STMT) return NULL; @@ -242,7 +246,10 @@ vect_recog_dot_prod_pattern (tree last_stmt, tree *type_in, tree *type_out) prod_type = half_type; stmt = SSA_NAME_DEF_STMT (oprnd0); - gcc_assert (stmt); + /* FORNOW. Can continue analyzing the def-use chain when this stmt in a phi + inside the loop (in case we are analyzing an outer-loop). */ + if (TREE_CODE (stmt) != GIMPLE_MODIFY_STMT) + return NULL; stmt_vinfo = vinfo_for_stmt (stmt); gcc_assert (stmt_vinfo); if (STMT_VINFO_DEF_TYPE (stmt_vinfo) != vect_loop_def) @@ -300,6 +307,16 @@ vect_recog_dot_prod_pattern (tree last_stmt, tree *type_in, tree *type_out) fprintf (vect_dump, "vect_recog_dot_prod_pattern: detected: "); print_generic_expr (vect_dump, pattern_expr, TDF_SLIM); } + + /* We don't allow changing the order of the computation in the inner-loop + when doing outer-loop vectorization. */ + if (nested_in_vect_loop_p (loop, last_stmt)) + { + if (vect_print_dump_info (REPORT_DETAILS)) + fprintf (vect_dump, "vect_recog_dot_prod_pattern: not allowed."); + return NULL; + } + return pattern_expr; } @@ -434,7 +451,7 @@ vect_recog_pow_pattern (tree last_stmt, tree *type_in, tree *type_out) { tree expr; tree type; - tree fn, arglist, base, exp; + tree fn, base, exp; if (TREE_CODE (last_stmt) != GIMPLE_MODIFY_STMT) return NULL; @@ -446,15 +463,14 @@ vect_recog_pow_pattern (tree last_stmt, tree *type_in, tree *type_out) return NULL_TREE; fn = get_callee_fndecl (expr); - arglist = TREE_OPERAND (expr, 1); switch (DECL_FUNCTION_CODE (fn)) { case BUILT_IN_POWIF: case BUILT_IN_POWI: case BUILT_IN_POWF: case BUILT_IN_POW: - base = TREE_VALUE (arglist); - exp = TREE_VALUE (TREE_CHAIN (arglist)); + base = CALL_EXPR_ARG (expr, 0); + exp = CALL_EXPR_ARG (expr, 1); if (TREE_CODE (exp) != REAL_CST && TREE_CODE (exp) != INTEGER_CST) return NULL_TREE; @@ -484,11 +500,10 @@ vect_recog_pow_pattern (tree last_stmt, tree *type_in, tree *type_out) && REAL_VALUES_EQUAL (TREE_REAL_CST (exp), dconsthalf)) { tree newfn = mathfn_built_in (TREE_TYPE (base), BUILT_IN_SQRT); - tree newarglist = build_tree_list (NULL_TREE, base); *type_in = get_vectype_for_scalar_type (TREE_TYPE (base)); if (*type_in) { - newfn = build_function_call_expr (newfn, newarglist); + newfn = build_call_expr (newfn, 1, base); if (vectorizable_function (newfn, *type_in, *type_in) != NULL_TREE) return newfn; } @@ -528,7 +543,14 @@ vect_recog_pow_pattern (tree last_stmt, tree *type_in, tree *type_out) * Return value: A new stmt that will be used to replace the sequence of stmts that constitute the pattern. In this case it will be: WIDEN_SUM -*/ + + Note: The widening-sum idiom is a widening reduction pattern that is + vectorized without preserving all the intermediate results. It + produces only N/2 (widened) results (by summing up pairs of + intermediate results) rather than all N results. Therefore, we + cannot allow this pattern when we want to get all the results and in + the correct order (as is the case when this computation is in an + inner-loop nested in an outer-loop that us being vectorized). */ static tree vect_recog_widen_sum_pattern (tree last_stmt, tree *type_in, tree *type_out) @@ -538,6 +560,8 @@ vect_recog_widen_sum_pattern (tree last_stmt, tree *type_in, tree *type_out) stmt_vec_info stmt_vinfo = vinfo_for_stmt (last_stmt); tree type, half_type; tree pattern_expr; + loop_vec_info loop_info = STMT_VINFO_LOOP_VINFO (stmt_vinfo); + struct loop *loop = LOOP_VINFO_LOOP (loop_info); if (TREE_CODE (last_stmt) != GIMPLE_MODIFY_STMT) return NULL; @@ -587,6 +611,16 @@ vect_recog_widen_sum_pattern (tree last_stmt, tree *type_in, tree *type_out) fprintf (vect_dump, "vect_recog_widen_sum_pattern: detected: "); print_generic_expr (vect_dump, pattern_expr, TDF_SLIM); } + + /* We don't allow changing the order of the computation in the inner-loop + when doing outer-loop vectorization. */ + if (nested_in_vect_loop_p (loop, last_stmt)) + { + if (vect_print_dump_info (REPORT_DETAILS)) + fprintf (vect_dump, "vect_recog_widen_sum_pattern: not allowed."); + return NULL; + } + return pattern_expr; } @@ -654,11 +688,12 @@ vect_pattern_recog_1 ( optab = optab_for_tree_code (TREE_CODE (pattern_expr), pattern_vectype); vec_mode = TYPE_MODE (pattern_vectype); if (!optab - || (icode = optab->handlers[(int) vec_mode].insn_code) == + || (icode = optab_handler (optab, vec_mode)->insn_code) == CODE_FOR_nothing || (type_out - && (insn_data[icode].operand[0].mode != - TYPE_MODE (get_vectype_for_scalar_type (type_out))))) + && (!get_vectype_for_scalar_type (type_out) + || (insn_data[icode].operand[0].mode != + TYPE_MODE (get_vectype_for_scalar_type (type_out)))))) return; } @@ -676,8 +711,7 @@ vect_pattern_recog_1 ( var = create_tmp_var (pattern_type, "patt"); add_referenced_var (var); var_name = make_ssa_name (var, NULL_TREE); - pattern_expr = build2 (GIMPLE_MODIFY_STMT, void_type_node, var_name, - pattern_expr); + pattern_expr = build_gimple_modify_stmt (var_name, pattern_expr); SSA_NAME_DEF_STMT (var_name) = pattern_expr; bsi_insert_before (&si, pattern_expr, BSI_SAME_STMT); ann = stmt_ann (pattern_expr);