2 Copyright (C) 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010
3 Free Software Foundation, Inc.
4 Contributed by Dorit Naishlos <dorit@il.ibm.com>
6 This file is part of GCC.
8 GCC is free software; you can redistribute it and/or modify it under
9 the terms of the GNU General Public License as published by the Free
10 Software Foundation; either version 3, or (at your option) any later
13 GCC is distributed in the hope that it will be useful, but WITHOUT ANY
14 WARRANTY; without even the implied warranty of MERCHANTABILITY or
15 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
18 You should have received a copy of the GNU General Public License
19 along with GCC; see the file COPYING3. If not see
20 <http://www.gnu.org/licenses/>. */
22 /* Loop and basic block vectorizer.
24 This file contains drivers for the three vectorizers:
25 (1) loop vectorizer (inter-iteration parallelism),
26 (2) loop-aware SLP (intra-iteration parallelism) (invoked by the loop
28 (3) BB vectorizer (out-of-loops), aka SLP
30 The rest of the vectorizer's code is organized as follows:
31 - tree-vect-loop.c - loop specific parts such as reductions, etc. These are
32 used by drivers (1) and (2).
33 - tree-vect-loop-manip.c - vectorizer's loop control-flow utilities, used by
35 - tree-vect-slp.c - BB vectorization specific analysis and transformation,
36 used by drivers (2) and (3).
37 - tree-vect-stmts.c - statements analysis and transformation (used by all).
38 - tree-vect-data-refs.c - vectorizer specific data-refs analysis and
39 manipulations (used by all).
40 - tree-vect-patterns.c - vectorizable code patterns detector (used by all)
42 Here's a poor attempt at illustrating that:
45 loop_vect() loop_aware_slp() slp_vect()
48 tree-vect-loop.c tree-vect-slp.c
53 tree-vect-stmts.c tree-vect-data-refs.c
60 #include "coretypes.h"
64 #include "tree-pretty-print.h"
65 #include "tree-flow.h"
66 #include "tree-dump.h"
68 #include "cfglayout.h"
69 #include "tree-vectorizer.h"
70 #include "tree-pass.h"
73 /* vect_dump will be set to stderr or dump_file if exist. */
76 /* vect_verbosity_level set to an invalid value
77 to mark that it's uninitialized. */
78 static enum vect_verbosity_levels vect_verbosity_level = MAX_VERBOSITY_LEVEL;
80 /* Loop or bb location. */
83 /* Vector mapping GIMPLE stmt to stmt_vec_info. */
84 VEC(vec_void_p,heap) *stmt_vec_info_vec;
88 /* Function vect_set_dump_settings.
90 Fix the verbosity level of the vectorizer if the
91 requested level was not set explicitly using the flag
92 -ftree-vectorizer-verbose=N.
93 Decide where to print the debugging information (dump_file/stderr).
94 If the user defined the verbosity level, but there is no dump file,
95 print to stderr, otherwise print to the dump file. */
98 vect_set_dump_settings (bool slp)
100 vect_dump = dump_file;
102 /* Check if the verbosity level was defined by the user: */
103 if (user_vect_verbosity_level != MAX_VERBOSITY_LEVEL)
105 vect_verbosity_level = user_vect_verbosity_level;
106 /* Ignore user defined verbosity if dump flags require higher level of
110 if (((dump_flags & TDF_DETAILS)
111 && vect_verbosity_level >= REPORT_DETAILS)
112 || ((dump_flags & TDF_STATS)
113 && vect_verbosity_level >= REPORT_UNVECTORIZED_LOCATIONS))
118 /* If there is no dump file, print to stderr in case of loop
127 /* User didn't specify verbosity level: */
128 if (dump_file && (dump_flags & TDF_DETAILS))
129 vect_verbosity_level = REPORT_DETAILS;
130 else if (dump_file && (dump_flags & TDF_STATS))
131 vect_verbosity_level = REPORT_UNVECTORIZED_LOCATIONS;
133 vect_verbosity_level = REPORT_NONE;
135 gcc_assert (dump_file || vect_verbosity_level == REPORT_NONE);
139 /* Function debug_loop_details.
141 For vectorization debug dumps. */
144 vect_print_dump_info (enum vect_verbosity_levels vl)
146 if (vl > vect_verbosity_level)
149 if (!current_function_decl || !vect_dump)
153 fprintf (vect_dump, "\n");
155 else if (vect_location == UNKNOWN_LOC)
156 fprintf (vect_dump, "\n%s:%d: note: ",
157 DECL_SOURCE_FILE (current_function_decl),
158 DECL_SOURCE_LINE (current_function_decl));
160 fprintf (vect_dump, "\n%s:%d: note: ",
161 LOC_FILE (vect_location), LOC_LINE (vect_location));
167 /* Function vectorize_loops.
169 Entry point to loop vectorization phase. */
172 vectorize_loops (void)
175 unsigned int num_vectorized_loops = 0;
176 unsigned int vect_loops_num;
180 vect_loops_num = number_of_loops ();
182 /* Bail out if there are no loops. */
183 if (vect_loops_num <= 1)
186 /* Fix the verbosity level if not defined explicitly by the user. */
187 vect_set_dump_settings (false);
189 init_stmt_vec_info_vec ();
191 /* ----------- Analyze loops. ----------- */
193 /* If some loop was duplicated, it gets bigger number
194 than all previously defined loops. This fact allows us to run
195 only over initial loops skipping newly generated ones. */
196 FOR_EACH_LOOP (li, loop, 0)
197 if (optimize_loop_nest_for_speed_p (loop))
199 loop_vec_info loop_vinfo;
201 vect_location = find_loop_location (loop);
202 loop_vinfo = vect_analyze_loop (loop);
203 loop->aux = loop_vinfo;
205 if (!loop_vinfo || !LOOP_VINFO_VECTORIZABLE_P (loop_vinfo))
208 vect_transform_loop (loop_vinfo);
209 num_vectorized_loops++;
212 vect_location = UNKNOWN_LOC;
214 statistics_counter_event (cfun, "Vectorized loops", num_vectorized_loops);
215 if (vect_print_dump_info (REPORT_UNVECTORIZED_LOCATIONS)
216 || (num_vectorized_loops > 0
217 && vect_print_dump_info (REPORT_VECTORIZED_LOCATIONS)))
218 fprintf (vect_dump, "vectorized %u loops in function.\n",
219 num_vectorized_loops);
221 /* ----------- Finalize. ----------- */
223 mark_sym_for_renaming (gimple_vop (cfun));
225 for (i = 1; i < vect_loops_num; i++)
227 loop_vec_info loop_vinfo;
232 loop_vinfo = (loop_vec_info) loop->aux;
233 destroy_loop_vec_info (loop_vinfo, true);
237 free_stmt_vec_info_vec ();
239 return num_vectorized_loops > 0 ? TODO_cleanup_cfg : 0;
243 /* Entry point to basic block SLP phase. */
246 execute_vect_slp (void)
250 /* Fix the verbosity level if not defined explicitly by the user. */
251 vect_set_dump_settings (true);
253 init_stmt_vec_info_vec ();
257 vect_location = find_bb_location (bb);
259 if (vect_slp_analyze_bb (bb))
261 vect_slp_transform_bb (bb);
263 if (vect_print_dump_info (REPORT_VECTORIZED_LOCATIONS))
264 fprintf (vect_dump, "basic block vectorized using SLP\n");
268 free_stmt_vec_info_vec ();
275 /* Apply SLP either if the vectorizer is on and the user didn't specify
276 whether to run SLP or not, or if the SLP flag was set by the user. */
277 return ((flag_tree_vectorize != 0 && flag_tree_slp_vectorize != 0)
278 || flag_tree_slp_vectorize == 1);
281 struct gimple_opt_pass pass_slp_vectorize =
286 gate_vect_slp, /* gate */
287 execute_vect_slp, /* execute */
290 0, /* static_pass_number */
291 TV_TREE_SLP_VECTORIZATION, /* tv_id */
292 PROP_ssa | PROP_cfg, /* properties_required */
293 0, /* properties_provided */
294 0, /* properties_destroyed */
295 0, /* todo_flags_start */
299 | TODO_verify_stmts /* todo_flags_finish */
304 /* Increase alignment of global arrays to improve vectorization potential.
306 - Consider also structs that have an array field.
307 - Use ipa analysis to prune arrays that can't be vectorized?
308 This should involve global alignment analysis and in the future also
312 increase_alignment (void)
314 struct varpool_node *vnode;
316 /* Increase the alignment of all global arrays for vectorization. */
317 for (vnode = varpool_nodes_queue;
319 vnode = vnode->next_needed)
321 tree vectype, decl = vnode->decl;
323 unsigned int alignment;
326 if (TREE_CODE (t) != ARRAY_TYPE)
328 vectype = get_vectype_for_scalar_type (strip_array_types (t));
331 alignment = TYPE_ALIGN (vectype);
332 if (DECL_ALIGN (decl) >= alignment)
335 if (vect_can_force_dr_alignment_p (decl, alignment))
337 DECL_ALIGN (decl) = TYPE_ALIGN (vectype);
338 DECL_USER_ALIGN (decl) = 1;
341 fprintf (dump_file, "Increasing alignment of decl: ");
342 print_generic_expr (dump_file, decl, TDF_SLIM);
343 fprintf (dump_file, "\n");
352 gate_increase_alignment (void)
354 return flag_section_anchors && flag_tree_vectorize;
358 struct simple_ipa_opt_pass pass_ipa_increase_alignment =
362 "increase_alignment", /* name */
363 gate_increase_alignment, /* gate */
364 increase_alignment, /* execute */
367 0, /* static_pass_number */
368 TV_IPA_OPT, /* tv_id */
369 0, /* properties_required */
370 0, /* properties_provided */
371 0, /* properties_destroyed */
372 0, /* todo_flags_start */
373 0 /* todo_flags_finish */