gcc50: Backport c++11 literal fixes from gcc80.
[dragonfly.git] / contrib / gcc-5.0 / gcc / ira.c
blob72e1043ba783702d26f8cad09c0ae78d04b47399
1 /* Integrated Register Allocator (IRA) entry point.
2 Copyright (C) 2006-2015 Free Software Foundation, Inc.
3 Contributed by Vladimir Makarov <vmakarov@redhat.com>.
5 This file is part of GCC.
7 GCC is free software; you can redistribute it and/or modify it under
8 the terms of the GNU General Public License as published by the Free
9 Software Foundation; either version 3, or (at your option) any later
10 version.
12 GCC is distributed in the hope that it will be useful, but WITHOUT ANY
13 WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
15 for more details.
17 You should have received a copy of the GNU General Public License
18 along with GCC; see the file COPYING3. If not see
19 <http://www.gnu.org/licenses/>. */
21 /* The integrated register allocator (IRA) is a
22 regional register allocator performing graph coloring on a top-down
23 traversal of nested regions. Graph coloring in a region is based
24 on Chaitin-Briggs algorithm. It is called integrated because
25 register coalescing, register live range splitting, and choosing a
26 better hard register are done on-the-fly during coloring. Register
27 coalescing and choosing a cheaper hard register is done by hard
28 register preferencing during hard register assigning. The live
29 range splitting is a byproduct of the regional register allocation.
31 Major IRA notions are:
33 o *Region* is a part of CFG where graph coloring based on
34 Chaitin-Briggs algorithm is done. IRA can work on any set of
35 nested CFG regions forming a tree. Currently the regions are
36 the entire function for the root region and natural loops for
37 the other regions. Therefore data structure representing a
38 region is called loop_tree_node.
40 o *Allocno class* is a register class used for allocation of
41 given allocno. It means that only hard register of given
42 register class can be assigned to given allocno. In reality,
43 even smaller subset of (*profitable*) hard registers can be
44 assigned. In rare cases, the subset can be even smaller
45 because our modification of Chaitin-Briggs algorithm requires
46 that sets of hard registers can be assigned to allocnos forms a
47 forest, i.e. the sets can be ordered in a way where any
48 previous set is not intersected with given set or is a superset
49 of given set.
51 o *Pressure class* is a register class belonging to a set of
52 register classes containing all of the hard-registers available
53 for register allocation. The set of all pressure classes for a
54 target is defined in the corresponding machine-description file
55 according some criteria. Register pressure is calculated only
56 for pressure classes and it affects some IRA decisions as
57 forming allocation regions.
59 o *Allocno* represents the live range of a pseudo-register in a
60 region. Besides the obvious attributes like the corresponding
61 pseudo-register number, allocno class, conflicting allocnos and
62 conflicting hard-registers, there are a few allocno attributes
63 which are important for understanding the allocation algorithm:
65 - *Live ranges*. This is a list of ranges of *program points*
66 where the allocno lives. Program points represent places
67 where a pseudo can be born or become dead (there are
68 approximately two times more program points than the insns)
69 and they are represented by integers starting with 0. The
70 live ranges are used to find conflicts between allocnos.
71 They also play very important role for the transformation of
72 the IRA internal representation of several regions into a one
73 region representation. The later is used during the reload
74 pass work because each allocno represents all of the
75 corresponding pseudo-registers.
77 - *Hard-register costs*. This is a vector of size equal to the
78 number of available hard-registers of the allocno class. The
79 cost of a callee-clobbered hard-register for an allocno is
80 increased by the cost of save/restore code around the calls
81 through the given allocno's life. If the allocno is a move
82 instruction operand and another operand is a hard-register of
83 the allocno class, the cost of the hard-register is decreased
84 by the move cost.
86 When an allocno is assigned, the hard-register with minimal
87 full cost is used. Initially, a hard-register's full cost is
88 the corresponding value from the hard-register's cost vector.
89 If the allocno is connected by a *copy* (see below) to
90 another allocno which has just received a hard-register, the
91 cost of the hard-register is decreased. Before choosing a
92 hard-register for an allocno, the allocno's current costs of
93 the hard-registers are modified by the conflict hard-register
94 costs of all of the conflicting allocnos which are not
95 assigned yet.
97 - *Conflict hard-register costs*. This is a vector of the same
98 size as the hard-register costs vector. To permit an
99 unassigned allocno to get a better hard-register, IRA uses
100 this vector to calculate the final full cost of the
101 available hard-registers. Conflict hard-register costs of an
102 unassigned allocno are also changed with a change of the
103 hard-register cost of the allocno when a copy involving the
104 allocno is processed as described above. This is done to
105 show other unassigned allocnos that a given allocno prefers
106 some hard-registers in order to remove the move instruction
107 corresponding to the copy.
109 o *Cap*. If a pseudo-register does not live in a region but
110 lives in a nested region, IRA creates a special allocno called
111 a cap in the outer region. A region cap is also created for a
112 subregion cap.
114 o *Copy*. Allocnos can be connected by copies. Copies are used
115 to modify hard-register costs for allocnos during coloring.
116 Such modifications reflects a preference to use the same
117 hard-register for the allocnos connected by copies. Usually
118 copies are created for move insns (in this case it results in
119 register coalescing). But IRA also creates copies for operands
120 of an insn which should be assigned to the same hard-register
121 due to constraints in the machine description (it usually
122 results in removing a move generated in reload to satisfy
123 the constraints) and copies referring to the allocno which is
124 the output operand of an instruction and the allocno which is
125 an input operand dying in the instruction (creation of such
126 copies results in less register shuffling). IRA *does not*
127 create copies between the same register allocnos from different
128 regions because we use another technique for propagating
129 hard-register preference on the borders of regions.
131 Allocnos (including caps) for the upper region in the region tree
132 *accumulate* information important for coloring from allocnos with
133 the same pseudo-register from nested regions. This includes
134 hard-register and memory costs, conflicts with hard-registers,
135 allocno conflicts, allocno copies and more. *Thus, attributes for
136 allocnos in a region have the same values as if the region had no
137 subregions*. It means that attributes for allocnos in the
138 outermost region corresponding to the function have the same values
139 as though the allocation used only one region which is the entire
140 function. It also means that we can look at IRA work as if the
141 first IRA did allocation for all function then it improved the
142 allocation for loops then their subloops and so on.
144 IRA major passes are:
146 o Building IRA internal representation which consists of the
147 following subpasses:
149 * First, IRA builds regions and creates allocnos (file
150 ira-build.c) and initializes most of their attributes.
152 * Then IRA finds an allocno class for each allocno and
153 calculates its initial (non-accumulated) cost of memory and
154 each hard-register of its allocno class (file ira-cost.c).
156 * IRA creates live ranges of each allocno, calculates register
157 pressure for each pressure class in each region, sets up
158 conflict hard registers for each allocno and info about calls
159 the allocno lives through (file ira-lives.c).
161 * IRA removes low register pressure loops from the regions
162 mostly to speed IRA up (file ira-build.c).
164 * IRA propagates accumulated allocno info from lower region
165 allocnos to corresponding upper region allocnos (file
166 ira-build.c).
168 * IRA creates all caps (file ira-build.c).
170 * Having live-ranges of allocnos and their classes, IRA creates
171 conflicting allocnos for each allocno. Conflicting allocnos
172 are stored as a bit vector or array of pointers to the
173 conflicting allocnos whatever is more profitable (file
174 ira-conflicts.c). At this point IRA creates allocno copies.
176 o Coloring. Now IRA has all necessary info to start graph coloring
177 process. It is done in each region on top-down traverse of the
178 region tree (file ira-color.c). There are following subpasses:
180 * Finding profitable hard registers of corresponding allocno
181 class for each allocno. For example, only callee-saved hard
182 registers are frequently profitable for allocnos living
183 through colors. If the profitable hard register set of
184 allocno does not form a tree based on subset relation, we use
185 some approximation to form the tree. This approximation is
186 used to figure out trivial colorability of allocnos. The
187 approximation is a pretty rare case.
189 * Putting allocnos onto the coloring stack. IRA uses Briggs
190 optimistic coloring which is a major improvement over
191 Chaitin's coloring. Therefore IRA does not spill allocnos at
192 this point. There is some freedom in the order of putting
193 allocnos on the stack which can affect the final result of
194 the allocation. IRA uses some heuristics to improve the
195 order. The major one is to form *threads* from colorable
196 allocnos and push them on the stack by threads. Thread is a
197 set of non-conflicting colorable allocnos connected by
198 copies. The thread contains allocnos from the colorable
199 bucket or colorable allocnos already pushed onto the coloring
200 stack. Pushing thread allocnos one after another onto the
201 stack increases chances of removing copies when the allocnos
202 get the same hard reg.
204 We also use a modification of Chaitin-Briggs algorithm which
205 works for intersected register classes of allocnos. To
206 figure out trivial colorability of allocnos, the mentioned
207 above tree of hard register sets is used. To get an idea how
208 the algorithm works in i386 example, let us consider an
209 allocno to which any general hard register can be assigned.
210 If the allocno conflicts with eight allocnos to which only
211 EAX register can be assigned, given allocno is still
212 trivially colorable because all conflicting allocnos might be
213 assigned only to EAX and all other general hard registers are
214 still free.
216 To get an idea of the used trivial colorability criterion, it
217 is also useful to read article "Graph-Coloring Register
218 Allocation for Irregular Architectures" by Michael D. Smith
219 and Glen Holloway. Major difference between the article
220 approach and approach used in IRA is that Smith's approach
221 takes register classes only from machine description and IRA
222 calculate register classes from intermediate code too
223 (e.g. an explicit usage of hard registers in RTL code for
224 parameter passing can result in creation of additional
225 register classes which contain or exclude the hard
226 registers). That makes IRA approach useful for improving
227 coloring even for architectures with regular register files
228 and in fact some benchmarking shows the improvement for
229 regular class architectures is even bigger than for irregular
230 ones. Another difference is that Smith's approach chooses
231 intersection of classes of all insn operands in which a given
232 pseudo occurs. IRA can use bigger classes if it is still
233 more profitable than memory usage.
235 * Popping the allocnos from the stack and assigning them hard
236 registers. If IRA can not assign a hard register to an
237 allocno and the allocno is coalesced, IRA undoes the
238 coalescing and puts the uncoalesced allocnos onto the stack in
239 the hope that some such allocnos will get a hard register
240 separately. If IRA fails to assign hard register or memory
241 is more profitable for it, IRA spills the allocno. IRA
242 assigns the allocno the hard-register with minimal full
243 allocation cost which reflects the cost of usage of the
244 hard-register for the allocno and cost of usage of the
245 hard-register for allocnos conflicting with given allocno.
247 * Chaitin-Briggs coloring assigns as many pseudos as possible
248 to hard registers. After coloring we try to improve
249 allocation with cost point of view. We improve the
250 allocation by spilling some allocnos and assigning the freed
251 hard registers to other allocnos if it decreases the overall
252 allocation cost.
254 * After allocno assigning in the region, IRA modifies the hard
255 register and memory costs for the corresponding allocnos in
256 the subregions to reflect the cost of possible loads, stores,
257 or moves on the border of the region and its subregions.
258 When default regional allocation algorithm is used
259 (-fira-algorithm=mixed), IRA just propagates the assignment
260 for allocnos if the register pressure in the region for the
261 corresponding pressure class is less than number of available
262 hard registers for given pressure class.
264 o Spill/restore code moving. When IRA performs an allocation
265 by traversing regions in top-down order, it does not know what
266 happens below in the region tree. Therefore, sometimes IRA
267 misses opportunities to perform a better allocation. A simple
268 optimization tries to improve allocation in a region having
269 subregions and containing in another region. If the
270 corresponding allocnos in the subregion are spilled, it spills
271 the region allocno if it is profitable. The optimization
272 implements a simple iterative algorithm performing profitable
273 transformations while they are still possible. It is fast in
274 practice, so there is no real need for a better time complexity
275 algorithm.
277 o Code change. After coloring, two allocnos representing the
278 same pseudo-register outside and inside a region respectively
279 may be assigned to different locations (hard-registers or
280 memory). In this case IRA creates and uses a new
281 pseudo-register inside the region and adds code to move allocno
282 values on the region's borders. This is done during top-down
283 traversal of the regions (file ira-emit.c). In some
284 complicated cases IRA can create a new allocno to move allocno
285 values (e.g. when a swap of values stored in two hard-registers
286 is needed). At this stage, the new allocno is marked as
287 spilled. IRA still creates the pseudo-register and the moves
288 on the region borders even when both allocnos were assigned to
289 the same hard-register. If the reload pass spills a
290 pseudo-register for some reason, the effect will be smaller
291 because another allocno will still be in the hard-register. In
292 most cases, this is better then spilling both allocnos. If
293 reload does not change the allocation for the two
294 pseudo-registers, the trivial move will be removed by
295 post-reload optimizations. IRA does not generate moves for
296 allocnos assigned to the same hard register when the default
297 regional allocation algorithm is used and the register pressure
298 in the region for the corresponding pressure class is less than
299 number of available hard registers for given pressure class.
300 IRA also does some optimizations to remove redundant stores and
301 to reduce code duplication on the region borders.
303 o Flattening internal representation. After changing code, IRA
304 transforms its internal representation for several regions into
305 one region representation (file ira-build.c). This process is
306 called IR flattening. Such process is more complicated than IR
307 rebuilding would be, but is much faster.
309 o After IR flattening, IRA tries to assign hard registers to all
310 spilled allocnos. This is implemented by a simple and fast
311 priority coloring algorithm (see function
312 ira_reassign_conflict_allocnos::ira-color.c). Here new allocnos
313 created during the code change pass can be assigned to hard
314 registers.
316 o At the end IRA calls the reload pass. The reload pass
317 communicates with IRA through several functions in file
318 ira-color.c to improve its decisions in
320 * sharing stack slots for the spilled pseudos based on IRA info
321 about pseudo-register conflicts.
323 * reassigning hard-registers to all spilled pseudos at the end
324 of each reload iteration.
326 * choosing a better hard-register to spill based on IRA info
327 about pseudo-register live ranges and the register pressure
328 in places where the pseudo-register lives.
330 IRA uses a lot of data representing the target processors. These
331 data are initialized in file ira.c.
333 If function has no loops (or the loops are ignored when
334 -fira-algorithm=CB is used), we have classic Chaitin-Briggs
335 coloring (only instead of separate pass of coalescing, we use hard
336 register preferencing). In such case, IRA works much faster
337 because many things are not made (like IR flattening, the
338 spill/restore optimization, and the code change).
340 Literature is worth to read for better understanding the code:
342 o Preston Briggs, Keith D. Cooper, Linda Torczon. Improvements to
343 Graph Coloring Register Allocation.
345 o David Callahan, Brian Koblenz. Register allocation via
346 hierarchical graph coloring.
348 o Keith Cooper, Anshuman Dasgupta, Jason Eckhardt. Revisiting Graph
349 Coloring Register Allocation: A Study of the Chaitin-Briggs and
350 Callahan-Koblenz Algorithms.
352 o Guei-Yuan Lueh, Thomas Gross, and Ali-Reza Adl-Tabatabai. Global
353 Register Allocation Based on Graph Fusion.
355 o Michael D. Smith and Glenn Holloway. Graph-Coloring Register
356 Allocation for Irregular Architectures
358 o Vladimir Makarov. The Integrated Register Allocator for GCC.
360 o Vladimir Makarov. The top-down register allocator for irregular
361 register file architectures.
366 #include "config.h"
367 #include "system.h"
368 #include "coretypes.h"
369 #include "tm.h"
370 #include "regs.h"
371 #include "hash-set.h"
372 #include "machmode.h"
373 #include "vec.h"
374 #include "double-int.h"
375 #include "input.h"
376 #include "alias.h"
377 #include "symtab.h"
378 #include "wide-int.h"
379 #include "inchash.h"
380 #include "tree.h"
381 #include "rtl.h"
382 #include "tm_p.h"
383 #include "target.h"
384 #include "flags.h"
385 #include "obstack.h"
386 #include "bitmap.h"
387 #include "hard-reg-set.h"
388 #include "predict.h"
389 #include "function.h"
390 #include "dominance.h"
391 #include "cfg.h"
392 #include "cfgrtl.h"
393 #include "cfgbuild.h"
394 #include "cfgcleanup.h"
395 #include "basic-block.h"
396 #include "df.h"
397 #include "hashtab.h"
398 #include "statistics.h"
399 #include "real.h"
400 #include "fixed-value.h"
401 #include "insn-config.h"
402 #include "expmed.h"
403 #include "dojump.h"
404 #include "explow.h"
405 #include "calls.h"
406 #include "emit-rtl.h"
407 #include "varasm.h"
408 #include "stmt.h"
409 #include "expr.h"
410 #include "recog.h"
411 #include "params.h"
412 #include "tree-pass.h"
413 #include "output.h"
414 #include "except.h"
415 #include "reload.h"
416 #include "diagnostic-core.h"
417 #include "ggc.h"
418 #include "ira-int.h"
419 #include "lra.h"
420 #include "dce.h"
421 #include "dbgcnt.h"
422 #include "rtl-iter.h"
423 #include "shrink-wrap.h"
425 struct target_ira default_target_ira;
426 struct target_ira_int default_target_ira_int;
427 #if SWITCHABLE_TARGET
428 struct target_ira *this_target_ira = &default_target_ira;
429 struct target_ira_int *this_target_ira_int = &default_target_ira_int;
430 #endif
432 /* A modified value of flag `-fira-verbose' used internally. */
433 int internal_flag_ira_verbose;
435 /* Dump file of the allocator if it is not NULL. */
436 FILE *ira_dump_file;
438 /* The number of elements in the following array. */
439 int ira_spilled_reg_stack_slots_num;
441 /* The following array contains info about spilled pseudo-registers
442 stack slots used in current function so far. */
443 struct ira_spilled_reg_stack_slot *ira_spilled_reg_stack_slots;
445 /* Correspondingly overall cost of the allocation, overall cost before
446 reload, cost of the allocnos assigned to hard-registers, cost of
447 the allocnos assigned to memory, cost of loads, stores and register
448 move insns generated for pseudo-register live range splitting (see
449 ira-emit.c). */
450 int64_t ira_overall_cost, overall_cost_before;
451 int64_t ira_reg_cost, ira_mem_cost;
452 int64_t ira_load_cost, ira_store_cost, ira_shuffle_cost;
453 int ira_move_loops_num, ira_additional_jumps_num;
455 /* All registers that can be eliminated. */
457 HARD_REG_SET eliminable_regset;
459 /* Value of max_reg_num () before IRA work start. This value helps
460 us to recognize a situation when new pseudos were created during
461 IRA work. */
462 static int max_regno_before_ira;
464 /* Temporary hard reg set used for a different calculation. */
465 static HARD_REG_SET temp_hard_regset;
467 #define last_mode_for_init_move_cost \
468 (this_target_ira_int->x_last_mode_for_init_move_cost)
471 /* The function sets up the map IRA_REG_MODE_HARD_REGSET. */
472 static void
473 setup_reg_mode_hard_regset (void)
475 int i, m, hard_regno;
477 for (m = 0; m < NUM_MACHINE_MODES; m++)
478 for (hard_regno = 0; hard_regno < FIRST_PSEUDO_REGISTER; hard_regno++)
480 CLEAR_HARD_REG_SET (ira_reg_mode_hard_regset[hard_regno][m]);
481 for (i = hard_regno_nregs[hard_regno][m] - 1; i >= 0; i--)
482 if (hard_regno + i < FIRST_PSEUDO_REGISTER)
483 SET_HARD_REG_BIT (ira_reg_mode_hard_regset[hard_regno][m],
484 hard_regno + i);
489 #define no_unit_alloc_regs \
490 (this_target_ira_int->x_no_unit_alloc_regs)
492 /* The function sets up the three arrays declared above. */
493 static void
494 setup_class_hard_regs (void)
496 int cl, i, hard_regno, n;
497 HARD_REG_SET processed_hard_reg_set;
499 ira_assert (SHRT_MAX >= FIRST_PSEUDO_REGISTER);
500 for (cl = (int) N_REG_CLASSES - 1; cl >= 0; cl--)
502 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
503 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
504 CLEAR_HARD_REG_SET (processed_hard_reg_set);
505 for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
507 ira_non_ordered_class_hard_regs[cl][i] = -1;
508 ira_class_hard_reg_index[cl][i] = -1;
510 for (n = 0, i = 0; i < FIRST_PSEUDO_REGISTER; i++)
512 #ifdef REG_ALLOC_ORDER
513 hard_regno = reg_alloc_order[i];
514 #else
515 hard_regno = i;
516 #endif
517 if (TEST_HARD_REG_BIT (processed_hard_reg_set, hard_regno))
518 continue;
519 SET_HARD_REG_BIT (processed_hard_reg_set, hard_regno);
520 if (! TEST_HARD_REG_BIT (temp_hard_regset, hard_regno))
521 ira_class_hard_reg_index[cl][hard_regno] = -1;
522 else
524 ira_class_hard_reg_index[cl][hard_regno] = n;
525 ira_class_hard_regs[cl][n++] = hard_regno;
528 ira_class_hard_regs_num[cl] = n;
529 for (n = 0, i = 0; i < FIRST_PSEUDO_REGISTER; i++)
530 if (TEST_HARD_REG_BIT (temp_hard_regset, i))
531 ira_non_ordered_class_hard_regs[cl][n++] = i;
532 ira_assert (ira_class_hard_regs_num[cl] == n);
536 /* Set up global variables defining info about hard registers for the
537 allocation. These depend on USE_HARD_FRAME_P whose TRUE value means
538 that we can use the hard frame pointer for the allocation. */
539 static void
540 setup_alloc_regs (bool use_hard_frame_p)
542 #ifdef ADJUST_REG_ALLOC_ORDER
543 ADJUST_REG_ALLOC_ORDER;
544 #endif
545 COPY_HARD_REG_SET (no_unit_alloc_regs, fixed_reg_set);
546 if (! use_hard_frame_p)
547 SET_HARD_REG_BIT (no_unit_alloc_regs, HARD_FRAME_POINTER_REGNUM);
548 setup_class_hard_regs ();
553 #define alloc_reg_class_subclasses \
554 (this_target_ira_int->x_alloc_reg_class_subclasses)
556 /* Initialize the table of subclasses of each reg class. */
557 static void
558 setup_reg_subclasses (void)
560 int i, j;
561 HARD_REG_SET temp_hard_regset2;
563 for (i = 0; i < N_REG_CLASSES; i++)
564 for (j = 0; j < N_REG_CLASSES; j++)
565 alloc_reg_class_subclasses[i][j] = LIM_REG_CLASSES;
567 for (i = 0; i < N_REG_CLASSES; i++)
569 if (i == (int) NO_REGS)
570 continue;
572 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[i]);
573 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
574 if (hard_reg_set_empty_p (temp_hard_regset))
575 continue;
576 for (j = 0; j < N_REG_CLASSES; j++)
577 if (i != j)
579 enum reg_class *p;
581 COPY_HARD_REG_SET (temp_hard_regset2, reg_class_contents[j]);
582 AND_COMPL_HARD_REG_SET (temp_hard_regset2, no_unit_alloc_regs);
583 if (! hard_reg_set_subset_p (temp_hard_regset,
584 temp_hard_regset2))
585 continue;
586 p = &alloc_reg_class_subclasses[j][0];
587 while (*p != LIM_REG_CLASSES) p++;
588 *p = (enum reg_class) i;
595 /* Set up IRA_MEMORY_MOVE_COST and IRA_MAX_MEMORY_MOVE_COST. */
596 static void
597 setup_class_subset_and_memory_move_costs (void)
599 int cl, cl2, mode, cost;
600 HARD_REG_SET temp_hard_regset2;
602 for (mode = 0; mode < MAX_MACHINE_MODE; mode++)
603 ira_memory_move_cost[mode][NO_REGS][0]
604 = ira_memory_move_cost[mode][NO_REGS][1] = SHRT_MAX;
605 for (cl = (int) N_REG_CLASSES - 1; cl >= 0; cl--)
607 if (cl != (int) NO_REGS)
608 for (mode = 0; mode < MAX_MACHINE_MODE; mode++)
610 ira_max_memory_move_cost[mode][cl][0]
611 = ira_memory_move_cost[mode][cl][0]
612 = memory_move_cost ((machine_mode) mode,
613 (reg_class_t) cl, false);
614 ira_max_memory_move_cost[mode][cl][1]
615 = ira_memory_move_cost[mode][cl][1]
616 = memory_move_cost ((machine_mode) mode,
617 (reg_class_t) cl, true);
618 /* Costs for NO_REGS are used in cost calculation on the
619 1st pass when the preferred register classes are not
620 known yet. In this case we take the best scenario. */
621 if (ira_memory_move_cost[mode][NO_REGS][0]
622 > ira_memory_move_cost[mode][cl][0])
623 ira_max_memory_move_cost[mode][NO_REGS][0]
624 = ira_memory_move_cost[mode][NO_REGS][0]
625 = ira_memory_move_cost[mode][cl][0];
626 if (ira_memory_move_cost[mode][NO_REGS][1]
627 > ira_memory_move_cost[mode][cl][1])
628 ira_max_memory_move_cost[mode][NO_REGS][1]
629 = ira_memory_move_cost[mode][NO_REGS][1]
630 = ira_memory_move_cost[mode][cl][1];
633 for (cl = (int) N_REG_CLASSES - 1; cl >= 0; cl--)
634 for (cl2 = (int) N_REG_CLASSES - 1; cl2 >= 0; cl2--)
636 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
637 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
638 COPY_HARD_REG_SET (temp_hard_regset2, reg_class_contents[cl2]);
639 AND_COMPL_HARD_REG_SET (temp_hard_regset2, no_unit_alloc_regs);
640 ira_class_subset_p[cl][cl2]
641 = hard_reg_set_subset_p (temp_hard_regset, temp_hard_regset2);
642 if (! hard_reg_set_empty_p (temp_hard_regset2)
643 && hard_reg_set_subset_p (reg_class_contents[cl2],
644 reg_class_contents[cl]))
645 for (mode = 0; mode < MAX_MACHINE_MODE; mode++)
647 cost = ira_memory_move_cost[mode][cl2][0];
648 if (cost > ira_max_memory_move_cost[mode][cl][0])
649 ira_max_memory_move_cost[mode][cl][0] = cost;
650 cost = ira_memory_move_cost[mode][cl2][1];
651 if (cost > ira_max_memory_move_cost[mode][cl][1])
652 ira_max_memory_move_cost[mode][cl][1] = cost;
655 for (cl = (int) N_REG_CLASSES - 1; cl >= 0; cl--)
656 for (mode = 0; mode < MAX_MACHINE_MODE; mode++)
658 ira_memory_move_cost[mode][cl][0]
659 = ira_max_memory_move_cost[mode][cl][0];
660 ira_memory_move_cost[mode][cl][1]
661 = ira_max_memory_move_cost[mode][cl][1];
663 setup_reg_subclasses ();
668 /* Define the following macro if allocation through malloc if
669 preferable. */
670 #define IRA_NO_OBSTACK
672 #ifndef IRA_NO_OBSTACK
673 /* Obstack used for storing all dynamic data (except bitmaps) of the
674 IRA. */
675 static struct obstack ira_obstack;
676 #endif
678 /* Obstack used for storing all bitmaps of the IRA. */
679 static struct bitmap_obstack ira_bitmap_obstack;
681 /* Allocate memory of size LEN for IRA data. */
682 void *
683 ira_allocate (size_t len)
685 void *res;
687 #ifndef IRA_NO_OBSTACK
688 res = obstack_alloc (&ira_obstack, len);
689 #else
690 res = xmalloc (len);
691 #endif
692 return res;
695 /* Free memory ADDR allocated for IRA data. */
696 void
697 ira_free (void *addr ATTRIBUTE_UNUSED)
699 #ifndef IRA_NO_OBSTACK
700 /* do nothing */
701 #else
702 free (addr);
703 #endif
707 /* Allocate and returns bitmap for IRA. */
708 bitmap
709 ira_allocate_bitmap (void)
711 return BITMAP_ALLOC (&ira_bitmap_obstack);
714 /* Free bitmap B allocated for IRA. */
715 void
716 ira_free_bitmap (bitmap b ATTRIBUTE_UNUSED)
718 /* do nothing */
723 /* Output information about allocation of all allocnos (except for
724 caps) into file F. */
725 void
726 ira_print_disposition (FILE *f)
728 int i, n, max_regno;
729 ira_allocno_t a;
730 basic_block bb;
732 fprintf (f, "Disposition:");
733 max_regno = max_reg_num ();
734 for (n = 0, i = FIRST_PSEUDO_REGISTER; i < max_regno; i++)
735 for (a = ira_regno_allocno_map[i];
736 a != NULL;
737 a = ALLOCNO_NEXT_REGNO_ALLOCNO (a))
739 if (n % 4 == 0)
740 fprintf (f, "\n");
741 n++;
742 fprintf (f, " %4d:r%-4d", ALLOCNO_NUM (a), ALLOCNO_REGNO (a));
743 if ((bb = ALLOCNO_LOOP_TREE_NODE (a)->bb) != NULL)
744 fprintf (f, "b%-3d", bb->index);
745 else
746 fprintf (f, "l%-3d", ALLOCNO_LOOP_TREE_NODE (a)->loop_num);
747 if (ALLOCNO_HARD_REGNO (a) >= 0)
748 fprintf (f, " %3d", ALLOCNO_HARD_REGNO (a));
749 else
750 fprintf (f, " mem");
752 fprintf (f, "\n");
755 /* Outputs information about allocation of all allocnos into
756 stderr. */
757 void
758 ira_debug_disposition (void)
760 ira_print_disposition (stderr);
765 /* Set up ira_stack_reg_pressure_class which is the biggest pressure
766 register class containing stack registers or NO_REGS if there are
767 no stack registers. To find this class, we iterate through all
768 register pressure classes and choose the first register pressure
769 class containing all the stack registers and having the biggest
770 size. */
771 static void
772 setup_stack_reg_pressure_class (void)
774 ira_stack_reg_pressure_class = NO_REGS;
775 #ifdef STACK_REGS
777 int i, best, size;
778 enum reg_class cl;
779 HARD_REG_SET temp_hard_regset2;
781 CLEAR_HARD_REG_SET (temp_hard_regset);
782 for (i = FIRST_STACK_REG; i <= LAST_STACK_REG; i++)
783 SET_HARD_REG_BIT (temp_hard_regset, i);
784 best = 0;
785 for (i = 0; i < ira_pressure_classes_num; i++)
787 cl = ira_pressure_classes[i];
788 COPY_HARD_REG_SET (temp_hard_regset2, temp_hard_regset);
789 AND_HARD_REG_SET (temp_hard_regset2, reg_class_contents[cl]);
790 size = hard_reg_set_size (temp_hard_regset2);
791 if (best < size)
793 best = size;
794 ira_stack_reg_pressure_class = cl;
798 #endif
801 /* Find pressure classes which are register classes for which we
802 calculate register pressure in IRA, register pressure sensitive
803 insn scheduling, and register pressure sensitive loop invariant
804 motion.
806 To make register pressure calculation easy, we always use
807 non-intersected register pressure classes. A move of hard
808 registers from one register pressure class is not more expensive
809 than load and store of the hard registers. Most likely an allocno
810 class will be a subset of a register pressure class and in many
811 cases a register pressure class. That makes usage of register
812 pressure classes a good approximation to find a high register
813 pressure. */
814 static void
815 setup_pressure_classes (void)
817 int cost, i, n, curr;
818 int cl, cl2;
819 enum reg_class pressure_classes[N_REG_CLASSES];
820 int m;
821 HARD_REG_SET temp_hard_regset2;
822 bool insert_p;
824 n = 0;
825 for (cl = 0; cl < N_REG_CLASSES; cl++)
827 if (ira_class_hard_regs_num[cl] == 0)
828 continue;
829 if (ira_class_hard_regs_num[cl] != 1
830 /* A register class without subclasses may contain a few
831 hard registers and movement between them is costly
832 (e.g. SPARC FPCC registers). We still should consider it
833 as a candidate for a pressure class. */
834 && alloc_reg_class_subclasses[cl][0] < cl)
836 /* Check that the moves between any hard registers of the
837 current class are not more expensive for a legal mode
838 than load/store of the hard registers of the current
839 class. Such class is a potential candidate to be a
840 register pressure class. */
841 for (m = 0; m < NUM_MACHINE_MODES; m++)
843 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
844 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
845 AND_COMPL_HARD_REG_SET (temp_hard_regset,
846 ira_prohibited_class_mode_regs[cl][m]);
847 if (hard_reg_set_empty_p (temp_hard_regset))
848 continue;
849 ira_init_register_move_cost_if_necessary ((machine_mode) m);
850 cost = ira_register_move_cost[m][cl][cl];
851 if (cost <= ira_max_memory_move_cost[m][cl][1]
852 || cost <= ira_max_memory_move_cost[m][cl][0])
853 break;
855 if (m >= NUM_MACHINE_MODES)
856 continue;
858 curr = 0;
859 insert_p = true;
860 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
861 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
862 /* Remove so far added pressure classes which are subset of the
863 current candidate class. Prefer GENERAL_REGS as a pressure
864 register class to another class containing the same
865 allocatable hard registers. We do this because machine
866 dependent cost hooks might give wrong costs for the latter
867 class but always give the right cost for the former class
868 (GENERAL_REGS). */
869 for (i = 0; i < n; i++)
871 cl2 = pressure_classes[i];
872 COPY_HARD_REG_SET (temp_hard_regset2, reg_class_contents[cl2]);
873 AND_COMPL_HARD_REG_SET (temp_hard_regset2, no_unit_alloc_regs);
874 if (hard_reg_set_subset_p (temp_hard_regset, temp_hard_regset2)
875 && (! hard_reg_set_equal_p (temp_hard_regset, temp_hard_regset2)
876 || cl2 == (int) GENERAL_REGS))
878 pressure_classes[curr++] = (enum reg_class) cl2;
879 insert_p = false;
880 continue;
882 if (hard_reg_set_subset_p (temp_hard_regset2, temp_hard_regset)
883 && (! hard_reg_set_equal_p (temp_hard_regset2, temp_hard_regset)
884 || cl == (int) GENERAL_REGS))
885 continue;
886 if (hard_reg_set_equal_p (temp_hard_regset2, temp_hard_regset))
887 insert_p = false;
888 pressure_classes[curr++] = (enum reg_class) cl2;
890 /* If the current candidate is a subset of a so far added
891 pressure class, don't add it to the list of the pressure
892 classes. */
893 if (insert_p)
894 pressure_classes[curr++] = (enum reg_class) cl;
895 n = curr;
897 #ifdef ENABLE_IRA_CHECKING
899 HARD_REG_SET ignore_hard_regs;
901 /* Check pressure classes correctness: here we check that hard
902 registers from all register pressure classes contains all hard
903 registers available for the allocation. */
904 CLEAR_HARD_REG_SET (temp_hard_regset);
905 CLEAR_HARD_REG_SET (temp_hard_regset2);
906 COPY_HARD_REG_SET (ignore_hard_regs, no_unit_alloc_regs);
907 for (cl = 0; cl < LIM_REG_CLASSES; cl++)
909 /* For some targets (like MIPS with MD_REGS), there are some
910 classes with hard registers available for allocation but
911 not able to hold value of any mode. */
912 for (m = 0; m < NUM_MACHINE_MODES; m++)
913 if (contains_reg_of_mode[cl][m])
914 break;
915 if (m >= NUM_MACHINE_MODES)
917 IOR_HARD_REG_SET (ignore_hard_regs, reg_class_contents[cl]);
918 continue;
920 for (i = 0; i < n; i++)
921 if ((int) pressure_classes[i] == cl)
922 break;
923 IOR_HARD_REG_SET (temp_hard_regset2, reg_class_contents[cl]);
924 if (i < n)
925 IOR_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
927 for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
928 /* Some targets (like SPARC with ICC reg) have allocatable regs
929 for which no reg class is defined. */
930 if (REGNO_REG_CLASS (i) == NO_REGS)
931 SET_HARD_REG_BIT (ignore_hard_regs, i);
932 AND_COMPL_HARD_REG_SET (temp_hard_regset, ignore_hard_regs);
933 AND_COMPL_HARD_REG_SET (temp_hard_regset2, ignore_hard_regs);
934 ira_assert (hard_reg_set_subset_p (temp_hard_regset2, temp_hard_regset));
936 #endif
937 ira_pressure_classes_num = 0;
938 for (i = 0; i < n; i++)
940 cl = (int) pressure_classes[i];
941 ira_reg_pressure_class_p[cl] = true;
942 ira_pressure_classes[ira_pressure_classes_num++] = (enum reg_class) cl;
944 setup_stack_reg_pressure_class ();
947 /* Set up IRA_UNIFORM_CLASS_P. Uniform class is a register class
948 whose register move cost between any registers of the class is the
949 same as for all its subclasses. We use the data to speed up the
950 2nd pass of calculations of allocno costs. */
951 static void
952 setup_uniform_class_p (void)
954 int i, cl, cl2, m;
956 for (cl = 0; cl < N_REG_CLASSES; cl++)
958 ira_uniform_class_p[cl] = false;
959 if (ira_class_hard_regs_num[cl] == 0)
960 continue;
961 /* We can not use alloc_reg_class_subclasses here because move
962 cost hooks does not take into account that some registers are
963 unavailable for the subtarget. E.g. for i686, INT_SSE_REGS
964 is element of alloc_reg_class_subclasses for GENERAL_REGS
965 because SSE regs are unavailable. */
966 for (i = 0; (cl2 = reg_class_subclasses[cl][i]) != LIM_REG_CLASSES; i++)
968 if (ira_class_hard_regs_num[cl2] == 0)
969 continue;
970 for (m = 0; m < NUM_MACHINE_MODES; m++)
971 if (contains_reg_of_mode[cl][m] && contains_reg_of_mode[cl2][m])
973 ira_init_register_move_cost_if_necessary ((machine_mode) m);
974 if (ira_register_move_cost[m][cl][cl]
975 != ira_register_move_cost[m][cl2][cl2])
976 break;
978 if (m < NUM_MACHINE_MODES)
979 break;
981 if (cl2 == LIM_REG_CLASSES)
982 ira_uniform_class_p[cl] = true;
986 /* Set up IRA_ALLOCNO_CLASSES, IRA_ALLOCNO_CLASSES_NUM,
987 IRA_IMPORTANT_CLASSES, and IRA_IMPORTANT_CLASSES_NUM.
989 Target may have many subtargets and not all target hard registers can
990 be used for allocation, e.g. x86 port in 32-bit mode can not use
991 hard registers introduced in x86-64 like r8-r15). Some classes
992 might have the same allocatable hard registers, e.g. INDEX_REGS
993 and GENERAL_REGS in x86 port in 32-bit mode. To decrease different
994 calculations efforts we introduce allocno classes which contain
995 unique non-empty sets of allocatable hard-registers.
997 Pseudo class cost calculation in ira-costs.c is very expensive.
998 Therefore we are trying to decrease number of classes involved in
999 such calculation. Register classes used in the cost calculation
1000 are called important classes. They are allocno classes and other
1001 non-empty classes whose allocatable hard register sets are inside
1002 of an allocno class hard register set. From the first sight, it
1003 looks like that they are just allocno classes. It is not true. In
1004 example of x86-port in 32-bit mode, allocno classes will contain
1005 GENERAL_REGS but not LEGACY_REGS (because allocatable hard
1006 registers are the same for the both classes). The important
1007 classes will contain GENERAL_REGS and LEGACY_REGS. It is done
1008 because a machine description insn constraint may refers for
1009 LEGACY_REGS and code in ira-costs.c is mostly base on investigation
1010 of the insn constraints. */
1011 static void
1012 setup_allocno_and_important_classes (void)
1014 int i, j, n, cl;
1015 bool set_p;
1016 HARD_REG_SET temp_hard_regset2;
1017 static enum reg_class classes[LIM_REG_CLASSES + 1];
1019 n = 0;
1020 /* Collect classes which contain unique sets of allocatable hard
1021 registers. Prefer GENERAL_REGS to other classes containing the
1022 same set of hard registers. */
1023 for (i = 0; i < LIM_REG_CLASSES; i++)
1025 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[i]);
1026 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
1027 for (j = 0; j < n; j++)
1029 cl = classes[j];
1030 COPY_HARD_REG_SET (temp_hard_regset2, reg_class_contents[cl]);
1031 AND_COMPL_HARD_REG_SET (temp_hard_regset2,
1032 no_unit_alloc_regs);
1033 if (hard_reg_set_equal_p (temp_hard_regset,
1034 temp_hard_regset2))
1035 break;
1037 if (j >= n)
1038 classes[n++] = (enum reg_class) i;
1039 else if (i == GENERAL_REGS)
1040 /* Prefer general regs. For i386 example, it means that
1041 we prefer GENERAL_REGS over INDEX_REGS or LEGACY_REGS
1042 (all of them consists of the same available hard
1043 registers). */
1044 classes[j] = (enum reg_class) i;
1046 classes[n] = LIM_REG_CLASSES;
1048 /* Set up classes which can be used for allocnos as classes
1049 containing non-empty unique sets of allocatable hard
1050 registers. */
1051 ira_allocno_classes_num = 0;
1052 for (i = 0; (cl = classes[i]) != LIM_REG_CLASSES; i++)
1053 if (ira_class_hard_regs_num[cl] > 0)
1054 ira_allocno_classes[ira_allocno_classes_num++] = (enum reg_class) cl;
1055 ira_important_classes_num = 0;
1056 /* Add non-allocno classes containing to non-empty set of
1057 allocatable hard regs. */
1058 for (cl = 0; cl < N_REG_CLASSES; cl++)
1059 if (ira_class_hard_regs_num[cl] > 0)
1061 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
1062 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
1063 set_p = false;
1064 for (j = 0; j < ira_allocno_classes_num; j++)
1066 COPY_HARD_REG_SET (temp_hard_regset2,
1067 reg_class_contents[ira_allocno_classes[j]]);
1068 AND_COMPL_HARD_REG_SET (temp_hard_regset2, no_unit_alloc_regs);
1069 if ((enum reg_class) cl == ira_allocno_classes[j])
1070 break;
1071 else if (hard_reg_set_subset_p (temp_hard_regset,
1072 temp_hard_regset2))
1073 set_p = true;
1075 if (set_p && j >= ira_allocno_classes_num)
1076 ira_important_classes[ira_important_classes_num++]
1077 = (enum reg_class) cl;
1079 /* Now add allocno classes to the important classes. */
1080 for (j = 0; j < ira_allocno_classes_num; j++)
1081 ira_important_classes[ira_important_classes_num++]
1082 = ira_allocno_classes[j];
1083 for (cl = 0; cl < N_REG_CLASSES; cl++)
1085 ira_reg_allocno_class_p[cl] = false;
1086 ira_reg_pressure_class_p[cl] = false;
1088 for (j = 0; j < ira_allocno_classes_num; j++)
1089 ira_reg_allocno_class_p[ira_allocno_classes[j]] = true;
1090 setup_pressure_classes ();
1091 setup_uniform_class_p ();
1094 /* Setup translation in CLASS_TRANSLATE of all classes into a class
1095 given by array CLASSES of length CLASSES_NUM. The function is used
1096 make translation any reg class to an allocno class or to an
1097 pressure class. This translation is necessary for some
1098 calculations when we can use only allocno or pressure classes and
1099 such translation represents an approximate representation of all
1100 classes.
1102 The translation in case when allocatable hard register set of a
1103 given class is subset of allocatable hard register set of a class
1104 in CLASSES is pretty simple. We use smallest classes from CLASSES
1105 containing a given class. If allocatable hard register set of a
1106 given class is not a subset of any corresponding set of a class
1107 from CLASSES, we use the cheapest (with load/store point of view)
1108 class from CLASSES whose set intersects with given class set. */
1109 static void
1110 setup_class_translate_array (enum reg_class *class_translate,
1111 int classes_num, enum reg_class *classes)
1113 int cl, mode;
1114 enum reg_class aclass, best_class, *cl_ptr;
1115 int i, cost, min_cost, best_cost;
1117 for (cl = 0; cl < N_REG_CLASSES; cl++)
1118 class_translate[cl] = NO_REGS;
1120 for (i = 0; i < classes_num; i++)
1122 aclass = classes[i];
1123 for (cl_ptr = &alloc_reg_class_subclasses[aclass][0];
1124 (cl = *cl_ptr) != LIM_REG_CLASSES;
1125 cl_ptr++)
1126 if (class_translate[cl] == NO_REGS)
1127 class_translate[cl] = aclass;
1128 class_translate[aclass] = aclass;
1130 /* For classes which are not fully covered by one of given classes
1131 (in other words covered by more one given class), use the
1132 cheapest class. */
1133 for (cl = 0; cl < N_REG_CLASSES; cl++)
1135 if (cl == NO_REGS || class_translate[cl] != NO_REGS)
1136 continue;
1137 best_class = NO_REGS;
1138 best_cost = INT_MAX;
1139 for (i = 0; i < classes_num; i++)
1141 aclass = classes[i];
1142 COPY_HARD_REG_SET (temp_hard_regset,
1143 reg_class_contents[aclass]);
1144 AND_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
1145 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
1146 if (! hard_reg_set_empty_p (temp_hard_regset))
1148 min_cost = INT_MAX;
1149 for (mode = 0; mode < MAX_MACHINE_MODE; mode++)
1151 cost = (ira_memory_move_cost[mode][aclass][0]
1152 + ira_memory_move_cost[mode][aclass][1]);
1153 if (min_cost > cost)
1154 min_cost = cost;
1156 if (best_class == NO_REGS || best_cost > min_cost)
1158 best_class = aclass;
1159 best_cost = min_cost;
1163 class_translate[cl] = best_class;
1167 /* Set up array IRA_ALLOCNO_CLASS_TRANSLATE and
1168 IRA_PRESSURE_CLASS_TRANSLATE. */
1169 static void
1170 setup_class_translate (void)
1172 setup_class_translate_array (ira_allocno_class_translate,
1173 ira_allocno_classes_num, ira_allocno_classes);
1174 setup_class_translate_array (ira_pressure_class_translate,
1175 ira_pressure_classes_num, ira_pressure_classes);
1178 /* Order numbers of allocno classes in original target allocno class
1179 array, -1 for non-allocno classes. */
1180 static int allocno_class_order[N_REG_CLASSES];
1182 /* The function used to sort the important classes. */
1183 static int
1184 comp_reg_classes_func (const void *v1p, const void *v2p)
1186 enum reg_class cl1 = *(const enum reg_class *) v1p;
1187 enum reg_class cl2 = *(const enum reg_class *) v2p;
1188 enum reg_class tcl1, tcl2;
1189 int diff;
1191 tcl1 = ira_allocno_class_translate[cl1];
1192 tcl2 = ira_allocno_class_translate[cl2];
1193 if (tcl1 != NO_REGS && tcl2 != NO_REGS
1194 && (diff = allocno_class_order[tcl1] - allocno_class_order[tcl2]) != 0)
1195 return diff;
1196 return (int) cl1 - (int) cl2;
1199 /* For correct work of function setup_reg_class_relation we need to
1200 reorder important classes according to the order of their allocno
1201 classes. It places important classes containing the same
1202 allocatable hard register set adjacent to each other and allocno
1203 class with the allocatable hard register set right after the other
1204 important classes with the same set.
1206 In example from comments of function
1207 setup_allocno_and_important_classes, it places LEGACY_REGS and
1208 GENERAL_REGS close to each other and GENERAL_REGS is after
1209 LEGACY_REGS. */
1210 static void
1211 reorder_important_classes (void)
1213 int i;
1215 for (i = 0; i < N_REG_CLASSES; i++)
1216 allocno_class_order[i] = -1;
1217 for (i = 0; i < ira_allocno_classes_num; i++)
1218 allocno_class_order[ira_allocno_classes[i]] = i;
1219 qsort (ira_important_classes, ira_important_classes_num,
1220 sizeof (enum reg_class), comp_reg_classes_func);
1221 for (i = 0; i < ira_important_classes_num; i++)
1222 ira_important_class_nums[ira_important_classes[i]] = i;
1225 /* Set up IRA_REG_CLASS_SUBUNION, IRA_REG_CLASS_SUPERUNION,
1226 IRA_REG_CLASS_SUPER_CLASSES, IRA_REG_CLASSES_INTERSECT, and
1227 IRA_REG_CLASSES_INTERSECT_P. For the meaning of the relations,
1228 please see corresponding comments in ira-int.h. */
1229 static void
1230 setup_reg_class_relations (void)
1232 int i, cl1, cl2, cl3;
1233 HARD_REG_SET intersection_set, union_set, temp_set2;
1234 bool important_class_p[N_REG_CLASSES];
1236 memset (important_class_p, 0, sizeof (important_class_p));
1237 for (i = 0; i < ira_important_classes_num; i++)
1238 important_class_p[ira_important_classes[i]] = true;
1239 for (cl1 = 0; cl1 < N_REG_CLASSES; cl1++)
1241 ira_reg_class_super_classes[cl1][0] = LIM_REG_CLASSES;
1242 for (cl2 = 0; cl2 < N_REG_CLASSES; cl2++)
1244 ira_reg_classes_intersect_p[cl1][cl2] = false;
1245 ira_reg_class_intersect[cl1][cl2] = NO_REGS;
1246 ira_reg_class_subset[cl1][cl2] = NO_REGS;
1247 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl1]);
1248 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
1249 COPY_HARD_REG_SET (temp_set2, reg_class_contents[cl2]);
1250 AND_COMPL_HARD_REG_SET (temp_set2, no_unit_alloc_regs);
1251 if (hard_reg_set_empty_p (temp_hard_regset)
1252 && hard_reg_set_empty_p (temp_set2))
1254 /* The both classes have no allocatable hard registers
1255 -- take all class hard registers into account and use
1256 reg_class_subunion and reg_class_superunion. */
1257 for (i = 0;; i++)
1259 cl3 = reg_class_subclasses[cl1][i];
1260 if (cl3 == LIM_REG_CLASSES)
1261 break;
1262 if (reg_class_subset_p (ira_reg_class_intersect[cl1][cl2],
1263 (enum reg_class) cl3))
1264 ira_reg_class_intersect[cl1][cl2] = (enum reg_class) cl3;
1266 ira_reg_class_subunion[cl1][cl2] = reg_class_subunion[cl1][cl2];
1267 ira_reg_class_superunion[cl1][cl2] = reg_class_superunion[cl1][cl2];
1268 continue;
1270 ira_reg_classes_intersect_p[cl1][cl2]
1271 = hard_reg_set_intersect_p (temp_hard_regset, temp_set2);
1272 if (important_class_p[cl1] && important_class_p[cl2]
1273 && hard_reg_set_subset_p (temp_hard_regset, temp_set2))
1275 /* CL1 and CL2 are important classes and CL1 allocatable
1276 hard register set is inside of CL2 allocatable hard
1277 registers -- make CL1 a superset of CL2. */
1278 enum reg_class *p;
1280 p = &ira_reg_class_super_classes[cl1][0];
1281 while (*p != LIM_REG_CLASSES)
1282 p++;
1283 *p++ = (enum reg_class) cl2;
1284 *p = LIM_REG_CLASSES;
1286 ira_reg_class_subunion[cl1][cl2] = NO_REGS;
1287 ira_reg_class_superunion[cl1][cl2] = NO_REGS;
1288 COPY_HARD_REG_SET (intersection_set, reg_class_contents[cl1]);
1289 AND_HARD_REG_SET (intersection_set, reg_class_contents[cl2]);
1290 AND_COMPL_HARD_REG_SET (intersection_set, no_unit_alloc_regs);
1291 COPY_HARD_REG_SET (union_set, reg_class_contents[cl1]);
1292 IOR_HARD_REG_SET (union_set, reg_class_contents[cl2]);
1293 AND_COMPL_HARD_REG_SET (union_set, no_unit_alloc_regs);
1294 for (cl3 = 0; cl3 < N_REG_CLASSES; cl3++)
1296 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl3]);
1297 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
1298 if (hard_reg_set_subset_p (temp_hard_regset, intersection_set))
1300 /* CL3 allocatable hard register set is inside of
1301 intersection of allocatable hard register sets
1302 of CL1 and CL2. */
1303 if (important_class_p[cl3])
1305 COPY_HARD_REG_SET
1306 (temp_set2,
1307 reg_class_contents
1308 [(int) ira_reg_class_intersect[cl1][cl2]]);
1309 AND_COMPL_HARD_REG_SET (temp_set2, no_unit_alloc_regs);
1310 if (! hard_reg_set_subset_p (temp_hard_regset, temp_set2)
1311 /* If the allocatable hard register sets are
1312 the same, prefer GENERAL_REGS or the
1313 smallest class for debugging
1314 purposes. */
1315 || (hard_reg_set_equal_p (temp_hard_regset, temp_set2)
1316 && (cl3 == GENERAL_REGS
1317 || ((ira_reg_class_intersect[cl1][cl2]
1318 != GENERAL_REGS)
1319 && hard_reg_set_subset_p
1320 (reg_class_contents[cl3],
1321 reg_class_contents
1322 [(int)
1323 ira_reg_class_intersect[cl1][cl2]])))))
1324 ira_reg_class_intersect[cl1][cl2] = (enum reg_class) cl3;
1326 COPY_HARD_REG_SET
1327 (temp_set2,
1328 reg_class_contents[(int) ira_reg_class_subset[cl1][cl2]]);
1329 AND_COMPL_HARD_REG_SET (temp_set2, no_unit_alloc_regs);
1330 if (! hard_reg_set_subset_p (temp_hard_regset, temp_set2)
1331 /* Ignore unavailable hard registers and prefer
1332 smallest class for debugging purposes. */
1333 || (hard_reg_set_equal_p (temp_hard_regset, temp_set2)
1334 && hard_reg_set_subset_p
1335 (reg_class_contents[cl3],
1336 reg_class_contents
1337 [(int) ira_reg_class_subset[cl1][cl2]])))
1338 ira_reg_class_subset[cl1][cl2] = (enum reg_class) cl3;
1340 if (important_class_p[cl3]
1341 && hard_reg_set_subset_p (temp_hard_regset, union_set))
1343 /* CL3 allocatable hard register set is inside of
1344 union of allocatable hard register sets of CL1
1345 and CL2. */
1346 COPY_HARD_REG_SET
1347 (temp_set2,
1348 reg_class_contents[(int) ira_reg_class_subunion[cl1][cl2]]);
1349 AND_COMPL_HARD_REG_SET (temp_set2, no_unit_alloc_regs);
1350 if (ira_reg_class_subunion[cl1][cl2] == NO_REGS
1351 || (hard_reg_set_subset_p (temp_set2, temp_hard_regset)
1353 && (! hard_reg_set_equal_p (temp_set2,
1354 temp_hard_regset)
1355 || cl3 == GENERAL_REGS
1356 /* If the allocatable hard register sets are the
1357 same, prefer GENERAL_REGS or the smallest
1358 class for debugging purposes. */
1359 || (ira_reg_class_subunion[cl1][cl2] != GENERAL_REGS
1360 && hard_reg_set_subset_p
1361 (reg_class_contents[cl3],
1362 reg_class_contents
1363 [(int) ira_reg_class_subunion[cl1][cl2]])))))
1364 ira_reg_class_subunion[cl1][cl2] = (enum reg_class) cl3;
1366 if (hard_reg_set_subset_p (union_set, temp_hard_regset))
1368 /* CL3 allocatable hard register set contains union
1369 of allocatable hard register sets of CL1 and
1370 CL2. */
1371 COPY_HARD_REG_SET
1372 (temp_set2,
1373 reg_class_contents[(int) ira_reg_class_superunion[cl1][cl2]]);
1374 AND_COMPL_HARD_REG_SET (temp_set2, no_unit_alloc_regs);
1375 if (ira_reg_class_superunion[cl1][cl2] == NO_REGS
1376 || (hard_reg_set_subset_p (temp_hard_regset, temp_set2)
1378 && (! hard_reg_set_equal_p (temp_set2,
1379 temp_hard_regset)
1380 || cl3 == GENERAL_REGS
1381 /* If the allocatable hard register sets are the
1382 same, prefer GENERAL_REGS or the smallest
1383 class for debugging purposes. */
1384 || (ira_reg_class_superunion[cl1][cl2] != GENERAL_REGS
1385 && hard_reg_set_subset_p
1386 (reg_class_contents[cl3],
1387 reg_class_contents
1388 [(int) ira_reg_class_superunion[cl1][cl2]])))))
1389 ira_reg_class_superunion[cl1][cl2] = (enum reg_class) cl3;
1396 /* Output all uniform and important classes into file F. */
1397 static void
1398 print_unform_and_important_classes (FILE *f)
1400 static const char *const reg_class_names[] = REG_CLASS_NAMES;
1401 int i, cl;
1403 fprintf (f, "Uniform classes:\n");
1404 for (cl = 0; cl < N_REG_CLASSES; cl++)
1405 if (ira_uniform_class_p[cl])
1406 fprintf (f, " %s", reg_class_names[cl]);
1407 fprintf (f, "\nImportant classes:\n");
1408 for (i = 0; i < ira_important_classes_num; i++)
1409 fprintf (f, " %s", reg_class_names[ira_important_classes[i]]);
1410 fprintf (f, "\n");
1413 /* Output all possible allocno or pressure classes and their
1414 translation map into file F. */
1415 static void
1416 print_translated_classes (FILE *f, bool pressure_p)
1418 int classes_num = (pressure_p
1419 ? ira_pressure_classes_num : ira_allocno_classes_num);
1420 enum reg_class *classes = (pressure_p
1421 ? ira_pressure_classes : ira_allocno_classes);
1422 enum reg_class *class_translate = (pressure_p
1423 ? ira_pressure_class_translate
1424 : ira_allocno_class_translate);
1425 static const char *const reg_class_names[] = REG_CLASS_NAMES;
1426 int i;
1428 fprintf (f, "%s classes:\n", pressure_p ? "Pressure" : "Allocno");
1429 for (i = 0; i < classes_num; i++)
1430 fprintf (f, " %s", reg_class_names[classes[i]]);
1431 fprintf (f, "\nClass translation:\n");
1432 for (i = 0; i < N_REG_CLASSES; i++)
1433 fprintf (f, " %s -> %s\n", reg_class_names[i],
1434 reg_class_names[class_translate[i]]);
1437 /* Output all possible allocno and translation classes and the
1438 translation maps into stderr. */
1439 void
1440 ira_debug_allocno_classes (void)
1442 print_unform_and_important_classes (stderr);
1443 print_translated_classes (stderr, false);
1444 print_translated_classes (stderr, true);
1447 /* Set up different arrays concerning class subsets, allocno and
1448 important classes. */
1449 static void
1450 find_reg_classes (void)
1452 setup_allocno_and_important_classes ();
1453 setup_class_translate ();
1454 reorder_important_classes ();
1455 setup_reg_class_relations ();
1460 /* Set up the array above. */
1461 static void
1462 setup_hard_regno_aclass (void)
1464 int i;
1466 for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
1468 #if 1
1469 ira_hard_regno_allocno_class[i]
1470 = (TEST_HARD_REG_BIT (no_unit_alloc_regs, i)
1471 ? NO_REGS
1472 : ira_allocno_class_translate[REGNO_REG_CLASS (i)]);
1473 #else
1474 int j;
1475 enum reg_class cl;
1476 ira_hard_regno_allocno_class[i] = NO_REGS;
1477 for (j = 0; j < ira_allocno_classes_num; j++)
1479 cl = ira_allocno_classes[j];
1480 if (ira_class_hard_reg_index[cl][i] >= 0)
1482 ira_hard_regno_allocno_class[i] = cl;
1483 break;
1486 #endif
1492 /* Form IRA_REG_CLASS_MAX_NREGS and IRA_REG_CLASS_MIN_NREGS maps. */
1493 static void
1494 setup_reg_class_nregs (void)
1496 int i, cl, cl2, m;
1498 for (m = 0; m < MAX_MACHINE_MODE; m++)
1500 for (cl = 0; cl < N_REG_CLASSES; cl++)
1501 ira_reg_class_max_nregs[cl][m]
1502 = ira_reg_class_min_nregs[cl][m]
1503 = targetm.class_max_nregs ((reg_class_t) cl, (machine_mode) m);
1504 for (cl = 0; cl < N_REG_CLASSES; cl++)
1505 for (i = 0;
1506 (cl2 = alloc_reg_class_subclasses[cl][i]) != LIM_REG_CLASSES;
1507 i++)
1508 if (ira_reg_class_min_nregs[cl2][m]
1509 < ira_reg_class_min_nregs[cl][m])
1510 ira_reg_class_min_nregs[cl][m] = ira_reg_class_min_nregs[cl2][m];
1516 /* Set up IRA_PROHIBITED_CLASS_MODE_REGS and IRA_CLASS_SINGLETON.
1517 This function is called once IRA_CLASS_HARD_REGS has been initialized. */
1518 static void
1519 setup_prohibited_class_mode_regs (void)
1521 int j, k, hard_regno, cl, last_hard_regno, count;
1523 for (cl = (int) N_REG_CLASSES - 1; cl >= 0; cl--)
1525 COPY_HARD_REG_SET (temp_hard_regset, reg_class_contents[cl]);
1526 AND_COMPL_HARD_REG_SET (temp_hard_regset, no_unit_alloc_regs);
1527 for (j = 0; j < NUM_MACHINE_MODES; j++)
1529 count = 0;
1530 last_hard_regno = -1;
1531 CLEAR_HARD_REG_SET (ira_prohibited_class_mode_regs[cl][j]);
1532 for (k = ira_class_hard_regs_num[cl] - 1; k >= 0; k--)
1534 hard_regno = ira_class_hard_regs[cl][k];
1535 if (! HARD_REGNO_MODE_OK (hard_regno, (machine_mode) j))
1536 SET_HARD_REG_BIT (ira_prohibited_class_mode_regs[cl][j],
1537 hard_regno);
1538 else if (in_hard_reg_set_p (temp_hard_regset,
1539 (machine_mode) j, hard_regno))
1541 last_hard_regno = hard_regno;
1542 count++;
1545 ira_class_singleton[cl][j] = (count == 1 ? last_hard_regno : -1);
1550 /* Clarify IRA_PROHIBITED_CLASS_MODE_REGS by excluding hard registers
1551 spanning from one register pressure class to another one. It is
1552 called after defining the pressure classes. */
1553 static void
1554 clarify_prohibited_class_mode_regs (void)
1556 int j, k, hard_regno, cl, pclass, nregs;
1558 for (cl = (int) N_REG_CLASSES - 1; cl >= 0; cl--)
1559 for (j = 0; j < NUM_MACHINE_MODES; j++)
1561 CLEAR_HARD_REG_SET (ira_useful_class_mode_regs[cl][j]);
1562 for (k = ira_class_hard_regs_num[cl] - 1; k >= 0; k--)
1564 hard_regno = ira_class_hard_regs[cl][k];
1565 if (TEST_HARD_REG_BIT (ira_prohibited_class_mode_regs[cl][j], hard_regno))
1566 continue;
1567 nregs = hard_regno_nregs[hard_regno][j];
1568 if (hard_regno + nregs > FIRST_PSEUDO_REGISTER)
1570 SET_HARD_REG_BIT (ira_prohibited_class_mode_regs[cl][j],
1571 hard_regno);
1572 continue;
1574 pclass = ira_pressure_class_translate[REGNO_REG_CLASS (hard_regno)];
1575 for (nregs-- ;nregs >= 0; nregs--)
1576 if (((enum reg_class) pclass
1577 != ira_pressure_class_translate[REGNO_REG_CLASS
1578 (hard_regno + nregs)]))
1580 SET_HARD_REG_BIT (ira_prohibited_class_mode_regs[cl][j],
1581 hard_regno);
1582 break;
1584 if (!TEST_HARD_REG_BIT (ira_prohibited_class_mode_regs[cl][j],
1585 hard_regno))
1586 add_to_hard_reg_set (&ira_useful_class_mode_regs[cl][j],
1587 (machine_mode) j, hard_regno);
1592 /* Allocate and initialize IRA_REGISTER_MOVE_COST, IRA_MAY_MOVE_IN_COST
1593 and IRA_MAY_MOVE_OUT_COST for MODE. */
1594 void
1595 ira_init_register_move_cost (machine_mode mode)
1597 static unsigned short last_move_cost[N_REG_CLASSES][N_REG_CLASSES];
1598 bool all_match = true;
1599 unsigned int cl1, cl2;
1601 ira_assert (ira_register_move_cost[mode] == NULL
1602 && ira_may_move_in_cost[mode] == NULL
1603 && ira_may_move_out_cost[mode] == NULL);
1604 ira_assert (have_regs_of_mode[mode]);
1605 for (cl1 = 0; cl1 < N_REG_CLASSES; cl1++)
1606 for (cl2 = 0; cl2 < N_REG_CLASSES; cl2++)
1608 int cost;
1609 if (!contains_reg_of_mode[cl1][mode]
1610 || !contains_reg_of_mode[cl2][mode])
1612 if ((ira_reg_class_max_nregs[cl1][mode]
1613 > ira_class_hard_regs_num[cl1])
1614 || (ira_reg_class_max_nregs[cl2][mode]
1615 > ira_class_hard_regs_num[cl2]))
1616 cost = 65535;
1617 else
1618 cost = (ira_memory_move_cost[mode][cl1][0]
1619 + ira_memory_move_cost[mode][cl2][1]) * 2;
1621 else
1623 cost = register_move_cost (mode, (enum reg_class) cl1,
1624 (enum reg_class) cl2);
1625 ira_assert (cost < 65535);
1627 all_match &= (last_move_cost[cl1][cl2] == cost);
1628 last_move_cost[cl1][cl2] = cost;
1630 if (all_match && last_mode_for_init_move_cost != -1)
1632 ira_register_move_cost[mode]
1633 = ira_register_move_cost[last_mode_for_init_move_cost];
1634 ira_may_move_in_cost[mode]
1635 = ira_may_move_in_cost[last_mode_for_init_move_cost];
1636 ira_may_move_out_cost[mode]
1637 = ira_may_move_out_cost[last_mode_for_init_move_cost];
1638 return;
1640 last_mode_for_init_move_cost = mode;
1641 ira_register_move_cost[mode] = XNEWVEC (move_table, N_REG_CLASSES);
1642 ira_may_move_in_cost[mode] = XNEWVEC (move_table, N_REG_CLASSES);
1643 ira_may_move_out_cost[mode] = XNEWVEC (move_table, N_REG_CLASSES);
1644 for (cl1 = 0; cl1 < N_REG_CLASSES; cl1++)
1645 for (cl2 = 0; cl2 < N_REG_CLASSES; cl2++)
1647 int cost;
1648 enum reg_class *p1, *p2;
1650 if (last_move_cost[cl1][cl2] == 65535)
1652 ira_register_move_cost[mode][cl1][cl2] = 65535;
1653 ira_may_move_in_cost[mode][cl1][cl2] = 65535;
1654 ira_may_move_out_cost[mode][cl1][cl2] = 65535;
1656 else
1658 cost = last_move_cost[cl1][cl2];
1660 for (p2 = &reg_class_subclasses[cl2][0];
1661 *p2 != LIM_REG_CLASSES; p2++)
1662 if (ira_class_hard_regs_num[*p2] > 0
1663 && (ira_reg_class_max_nregs[*p2][mode]
1664 <= ira_class_hard_regs_num[*p2]))
1665 cost = MAX (cost, ira_register_move_cost[mode][cl1][*p2]);
1667 for (p1 = &reg_class_subclasses[cl1][0];
1668 *p1 != LIM_REG_CLASSES; p1++)
1669 if (ira_class_hard_regs_num[*p1] > 0
1670 && (ira_reg_class_max_nregs[*p1][mode]
1671 <= ira_class_hard_regs_num[*p1]))
1672 cost = MAX (cost, ira_register_move_cost[mode][*p1][cl2]);
1674 ira_assert (cost <= 65535);
1675 ira_register_move_cost[mode][cl1][cl2] = cost;
1677 if (ira_class_subset_p[cl1][cl2])
1678 ira_may_move_in_cost[mode][cl1][cl2] = 0;
1679 else
1680 ira_may_move_in_cost[mode][cl1][cl2] = cost;
1682 if (ira_class_subset_p[cl2][cl1])
1683 ira_may_move_out_cost[mode][cl1][cl2] = 0;
1684 else
1685 ira_may_move_out_cost[mode][cl1][cl2] = cost;
1692 /* This is called once during compiler work. It sets up
1693 different arrays whose values don't depend on the compiled
1694 function. */
1695 void
1696 ira_init_once (void)
1698 ira_init_costs_once ();
1699 lra_init_once ();
1702 /* Free ira_max_register_move_cost, ira_may_move_in_cost and
1703 ira_may_move_out_cost for each mode. */
1704 void
1705 target_ira_int::free_register_move_costs (void)
1707 int mode, i;
1709 /* Reset move_cost and friends, making sure we only free shared
1710 table entries once. */
1711 for (mode = 0; mode < MAX_MACHINE_MODE; mode++)
1712 if (x_ira_register_move_cost[mode])
1714 for (i = 0;
1715 i < mode && (x_ira_register_move_cost[i]
1716 != x_ira_register_move_cost[mode]);
1717 i++)
1719 if (i == mode)
1721 free (x_ira_register_move_cost[mode]);
1722 free (x_ira_may_move_in_cost[mode]);
1723 free (x_ira_may_move_out_cost[mode]);
1726 memset (x_ira_register_move_cost, 0, sizeof x_ira_register_move_cost);
1727 memset (x_ira_may_move_in_cost, 0, sizeof x_ira_may_move_in_cost);
1728 memset (x_ira_may_move_out_cost, 0, sizeof x_ira_may_move_out_cost);
1729 last_mode_for_init_move_cost = -1;
1732 target_ira_int::~target_ira_int ()
1734 free_ira_costs ();
1735 free_register_move_costs ();
1738 /* This is called every time when register related information is
1739 changed. */
1740 void
1741 ira_init (void)
1743 this_target_ira_int->free_register_move_costs ();
1744 setup_reg_mode_hard_regset ();
1745 setup_alloc_regs (flag_omit_frame_pointer != 0);
1746 setup_class_subset_and_memory_move_costs ();
1747 setup_reg_class_nregs ();
1748 setup_prohibited_class_mode_regs ();
1749 find_reg_classes ();
1750 clarify_prohibited_class_mode_regs ();
1751 setup_hard_regno_aclass ();
1752 ira_init_costs ();
1756 #define ira_prohibited_mode_move_regs_initialized_p \
1757 (this_target_ira_int->x_ira_prohibited_mode_move_regs_initialized_p)
1759 /* Set up IRA_PROHIBITED_MODE_MOVE_REGS. */
1760 static void
1761 setup_prohibited_mode_move_regs (void)
1763 int i, j;
1764 rtx test_reg1, test_reg2, move_pat;
1765 rtx_insn *move_insn;
1767 if (ira_prohibited_mode_move_regs_initialized_p)
1768 return;
1769 ira_prohibited_mode_move_regs_initialized_p = true;
1770 test_reg1 = gen_rtx_REG (VOIDmode, 0);
1771 test_reg2 = gen_rtx_REG (VOIDmode, 0);
1772 move_pat = gen_rtx_SET (VOIDmode, test_reg1, test_reg2);
1773 move_insn = gen_rtx_INSN (VOIDmode, 0, 0, 0, move_pat, 0, -1, 0);
1774 for (i = 0; i < NUM_MACHINE_MODES; i++)
1776 SET_HARD_REG_SET (ira_prohibited_mode_move_regs[i]);
1777 for (j = 0; j < FIRST_PSEUDO_REGISTER; j++)
1779 if (! HARD_REGNO_MODE_OK (j, (machine_mode) i))
1780 continue;
1781 SET_REGNO_RAW (test_reg1, j);
1782 PUT_MODE (test_reg1, (machine_mode) i);
1783 SET_REGNO_RAW (test_reg2, j);
1784 PUT_MODE (test_reg2, (machine_mode) i);
1785 INSN_CODE (move_insn) = -1;
1786 recog_memoized (move_insn);
1787 if (INSN_CODE (move_insn) < 0)
1788 continue;
1789 extract_insn (move_insn);
1790 /* We don't know whether the move will be in code that is optimized
1791 for size or speed, so consider all enabled alternatives. */
1792 if (! constrain_operands (1, get_enabled_alternatives (move_insn)))
1793 continue;
1794 CLEAR_HARD_REG_BIT (ira_prohibited_mode_move_regs[i], j);
1801 /* Setup possible alternatives in ALTS for INSN. */
1802 void
1803 ira_setup_alts (rtx_insn *insn, HARD_REG_SET &alts)
1805 /* MAP nalt * nop -> start of constraints for given operand and
1806 alternative. */
1807 static vec<const char *> insn_constraints;
1808 int nop, nalt;
1809 bool curr_swapped;
1810 const char *p;
1811 rtx op;
1812 int commutative = -1;
1814 extract_insn (insn);
1815 alternative_mask preferred = get_preferred_alternatives (insn);
1816 CLEAR_HARD_REG_SET (alts);
1817 insn_constraints.release ();
1818 insn_constraints.safe_grow_cleared (recog_data.n_operands
1819 * recog_data.n_alternatives + 1);
1820 /* Check that the hard reg set is enough for holding all
1821 alternatives. It is hard to imagine the situation when the
1822 assertion is wrong. */
1823 ira_assert (recog_data.n_alternatives
1824 <= (int) MAX (sizeof (HARD_REG_ELT_TYPE) * CHAR_BIT,
1825 FIRST_PSEUDO_REGISTER));
1826 for (curr_swapped = false;; curr_swapped = true)
1828 /* Calculate some data common for all alternatives to speed up the
1829 function. */
1830 for (nop = 0; nop < recog_data.n_operands; nop++)
1832 for (nalt = 0, p = recog_data.constraints[nop];
1833 nalt < recog_data.n_alternatives;
1834 nalt++)
1836 insn_constraints[nop * recog_data.n_alternatives + nalt] = p;
1837 while (*p && *p != ',')
1838 p++;
1839 if (*p)
1840 p++;
1843 for (nalt = 0; nalt < recog_data.n_alternatives; nalt++)
1845 if (!TEST_BIT (preferred, nalt)
1846 || TEST_HARD_REG_BIT (alts, nalt))
1847 continue;
1849 for (nop = 0; nop < recog_data.n_operands; nop++)
1851 int c, len;
1853 op = recog_data.operand[nop];
1854 p = insn_constraints[nop * recog_data.n_alternatives + nalt];
1855 if (*p == 0 || *p == ',')
1856 continue;
1859 switch (c = *p, len = CONSTRAINT_LEN (c, p), c)
1861 case '#':
1862 case ',':
1863 c = '\0';
1864 case '\0':
1865 len = 0;
1866 break;
1868 case '%':
1869 /* We only support one commutative marker, the
1870 first one. We already set commutative
1871 above. */
1872 if (commutative < 0)
1873 commutative = nop;
1874 break;
1876 case '0': case '1': case '2': case '3': case '4':
1877 case '5': case '6': case '7': case '8': case '9':
1878 goto op_success;
1879 break;
1881 case 'g':
1882 goto op_success;
1883 break;
1885 default:
1887 enum constraint_num cn = lookup_constraint (p);
1888 switch (get_constraint_type (cn))
1890 case CT_REGISTER:
1891 if (reg_class_for_constraint (cn) != NO_REGS)
1892 goto op_success;
1893 break;
1895 case CT_CONST_INT:
1896 if (CONST_INT_P (op)
1897 && (insn_const_int_ok_for_constraint
1898 (INTVAL (op), cn)))
1899 goto op_success;
1900 break;
1902 case CT_ADDRESS:
1903 case CT_MEMORY:
1904 goto op_success;
1906 case CT_FIXED_FORM:
1907 if (constraint_satisfied_p (op, cn))
1908 goto op_success;
1909 break;
1911 break;
1914 while (p += len, c);
1915 break;
1916 op_success:
1919 if (nop >= recog_data.n_operands)
1920 SET_HARD_REG_BIT (alts, nalt);
1922 if (commutative < 0)
1923 break;
1924 if (curr_swapped)
1925 break;
1926 op = recog_data.operand[commutative];
1927 recog_data.operand[commutative] = recog_data.operand[commutative + 1];
1928 recog_data.operand[commutative + 1] = op;
1933 /* Return the number of the output non-early clobber operand which
1934 should be the same in any case as operand with number OP_NUM (or
1935 negative value if there is no such operand). The function takes
1936 only really possible alternatives into consideration. */
1938 ira_get_dup_out_num (int op_num, HARD_REG_SET &alts)
1940 int curr_alt, c, original, dup;
1941 bool ignore_p, use_commut_op_p;
1942 const char *str;
1944 if (op_num < 0 || recog_data.n_alternatives == 0)
1945 return -1;
1946 /* We should find duplications only for input operands. */
1947 if (recog_data.operand_type[op_num] != OP_IN)
1948 return -1;
1949 str = recog_data.constraints[op_num];
1950 use_commut_op_p = false;
1951 for (;;)
1953 rtx op = recog_data.operand[op_num];
1955 for (curr_alt = 0, ignore_p = !TEST_HARD_REG_BIT (alts, curr_alt),
1956 original = -1;;)
1958 c = *str;
1959 if (c == '\0')
1960 break;
1961 if (c == '#')
1962 ignore_p = true;
1963 else if (c == ',')
1965 curr_alt++;
1966 ignore_p = !TEST_HARD_REG_BIT (alts, curr_alt);
1968 else if (! ignore_p)
1969 switch (c)
1971 case 'g':
1972 goto fail;
1973 default:
1975 enum constraint_num cn = lookup_constraint (str);
1976 enum reg_class cl = reg_class_for_constraint (cn);
1977 if (cl != NO_REGS
1978 && !targetm.class_likely_spilled_p (cl))
1979 goto fail;
1980 if (constraint_satisfied_p (op, cn))
1981 goto fail;
1982 break;
1985 case '0': case '1': case '2': case '3': case '4':
1986 case '5': case '6': case '7': case '8': case '9':
1987 if (original != -1 && original != c)
1988 goto fail;
1989 original = c;
1990 break;
1992 str += CONSTRAINT_LEN (c, str);
1994 if (original == -1)
1995 goto fail;
1996 dup = -1;
1997 for (ignore_p = false, str = recog_data.constraints[original - '0'];
1998 *str != 0;
1999 str++)
2000 if (ignore_p)
2002 if (*str == ',')
2003 ignore_p = false;
2005 else if (*str == '#')
2006 ignore_p = true;
2007 else if (! ignore_p)
2009 if (*str == '=')
2010 dup = original - '0';
2011 /* It is better ignore an alternative with early clobber. */
2012 else if (*str == '&')
2013 goto fail;
2015 if (dup >= 0)
2016 return dup;
2017 fail:
2018 if (use_commut_op_p)
2019 break;
2020 use_commut_op_p = true;
2021 if (recog_data.constraints[op_num][0] == '%')
2022 str = recog_data.constraints[op_num + 1];
2023 else if (op_num > 0 && recog_data.constraints[op_num - 1][0] == '%')
2024 str = recog_data.constraints[op_num - 1];
2025 else
2026 break;
2028 return -1;
2033 /* Search forward to see if the source register of a copy insn dies
2034 before either it or the destination register is modified, but don't
2035 scan past the end of the basic block. If so, we can replace the
2036 source with the destination and let the source die in the copy
2037 insn.
2039 This will reduce the number of registers live in that range and may
2040 enable the destination and the source coalescing, thus often saving
2041 one register in addition to a register-register copy. */
2043 static void
2044 decrease_live_ranges_number (void)
2046 basic_block bb;
2047 rtx_insn *insn;
2048 rtx set, src, dest, dest_death, q, note;
2049 rtx_insn *p;
2050 int sregno, dregno;
2052 if (! flag_expensive_optimizations)
2053 return;
2055 if (ira_dump_file)
2056 fprintf (ira_dump_file, "Starting decreasing number of live ranges...\n");
2058 FOR_EACH_BB_FN (bb, cfun)
2059 FOR_BB_INSNS (bb, insn)
2061 set = single_set (insn);
2062 if (! set)
2063 continue;
2064 src = SET_SRC (set);
2065 dest = SET_DEST (set);
2066 if (! REG_P (src) || ! REG_P (dest)
2067 || find_reg_note (insn, REG_DEAD, src))
2068 continue;
2069 sregno = REGNO (src);
2070 dregno = REGNO (dest);
2072 /* We don't want to mess with hard regs if register classes
2073 are small. */
2074 if (sregno == dregno
2075 || (targetm.small_register_classes_for_mode_p (GET_MODE (src))
2076 && (sregno < FIRST_PSEUDO_REGISTER
2077 || dregno < FIRST_PSEUDO_REGISTER))
2078 /* We don't see all updates to SP if they are in an
2079 auto-inc memory reference, so we must disallow this
2080 optimization on them. */
2081 || sregno == STACK_POINTER_REGNUM
2082 || dregno == STACK_POINTER_REGNUM)
2083 continue;
2085 dest_death = NULL_RTX;
2087 for (p = NEXT_INSN (insn); p; p = NEXT_INSN (p))
2089 if (! INSN_P (p))
2090 continue;
2091 if (BLOCK_FOR_INSN (p) != bb)
2092 break;
2094 if (reg_set_p (src, p) || reg_set_p (dest, p)
2095 /* If SRC is an asm-declared register, it must not be
2096 replaced in any asm. Unfortunately, the REG_EXPR
2097 tree for the asm variable may be absent in the SRC
2098 rtx, so we can't check the actual register
2099 declaration easily (the asm operand will have it,
2100 though). To avoid complicating the test for a rare
2101 case, we just don't perform register replacement
2102 for a hard reg mentioned in an asm. */
2103 || (sregno < FIRST_PSEUDO_REGISTER
2104 && asm_noperands (PATTERN (p)) >= 0
2105 && reg_overlap_mentioned_p (src, PATTERN (p)))
2106 /* Don't change hard registers used by a call. */
2107 || (CALL_P (p) && sregno < FIRST_PSEUDO_REGISTER
2108 && find_reg_fusage (p, USE, src))
2109 /* Don't change a USE of a register. */
2110 || (GET_CODE (PATTERN (p)) == USE
2111 && reg_overlap_mentioned_p (src, XEXP (PATTERN (p), 0))))
2112 break;
2114 /* See if all of SRC dies in P. This test is slightly
2115 more conservative than it needs to be. */
2116 if ((note = find_regno_note (p, REG_DEAD, sregno))
2117 && GET_MODE (XEXP (note, 0)) == GET_MODE (src))
2119 int failed = 0;
2121 /* We can do the optimization. Scan forward from INSN
2122 again, replacing regs as we go. Set FAILED if a
2123 replacement can't be done. In that case, we can't
2124 move the death note for SRC. This should be
2125 rare. */
2127 /* Set to stop at next insn. */
2128 for (q = next_real_insn (insn);
2129 q != next_real_insn (p);
2130 q = next_real_insn (q))
2132 if (reg_overlap_mentioned_p (src, PATTERN (q)))
2134 /* If SRC is a hard register, we might miss
2135 some overlapping registers with
2136 validate_replace_rtx, so we would have to
2137 undo it. We can't if DEST is present in
2138 the insn, so fail in that combination of
2139 cases. */
2140 if (sregno < FIRST_PSEUDO_REGISTER
2141 && reg_mentioned_p (dest, PATTERN (q)))
2142 failed = 1;
2144 /* Attempt to replace all uses. */
2145 else if (!validate_replace_rtx (src, dest, q))
2146 failed = 1;
2148 /* If this succeeded, but some part of the
2149 register is still present, undo the
2150 replacement. */
2151 else if (sregno < FIRST_PSEUDO_REGISTER
2152 && reg_overlap_mentioned_p (src, PATTERN (q)))
2154 validate_replace_rtx (dest, src, q);
2155 failed = 1;
2159 /* If DEST dies here, remove the death note and
2160 save it for later. Make sure ALL of DEST dies
2161 here; again, this is overly conservative. */
2162 if (! dest_death
2163 && (dest_death = find_regno_note (q, REG_DEAD, dregno)))
2165 if (GET_MODE (XEXP (dest_death, 0)) == GET_MODE (dest))
2166 remove_note (q, dest_death);
2167 else
2169 failed = 1;
2170 dest_death = 0;
2175 if (! failed)
2177 /* Move death note of SRC from P to INSN. */
2178 remove_note (p, note);
2179 XEXP (note, 1) = REG_NOTES (insn);
2180 REG_NOTES (insn) = note;
2183 /* DEST is also dead if INSN has a REG_UNUSED note for
2184 DEST. */
2185 if (! dest_death
2186 && (dest_death
2187 = find_regno_note (insn, REG_UNUSED, dregno)))
2189 PUT_REG_NOTE_KIND (dest_death, REG_DEAD);
2190 remove_note (insn, dest_death);
2193 /* Put death note of DEST on P if we saw it die. */
2194 if (dest_death)
2196 XEXP (dest_death, 1) = REG_NOTES (p);
2197 REG_NOTES (p) = dest_death;
2199 break;
2202 /* If SRC is a hard register which is set or killed in
2203 some other way, we can't do this optimization. */
2204 else if (sregno < FIRST_PSEUDO_REGISTER && dead_or_set_p (p, src))
2205 break;
2212 /* Return nonzero if REGNO is a particularly bad choice for reloading X. */
2213 static bool
2214 ira_bad_reload_regno_1 (int regno, rtx x)
2216 int x_regno, n, i;
2217 ira_allocno_t a;
2218 enum reg_class pref;
2220 /* We only deal with pseudo regs. */
2221 if (! x || GET_CODE (x) != REG)
2222 return false;
2224 x_regno = REGNO (x);
2225 if (x_regno < FIRST_PSEUDO_REGISTER)
2226 return false;
2228 /* If the pseudo prefers REGNO explicitly, then do not consider
2229 REGNO a bad spill choice. */
2230 pref = reg_preferred_class (x_regno);
2231 if (reg_class_size[pref] == 1)
2232 return !TEST_HARD_REG_BIT (reg_class_contents[pref], regno);
2234 /* If the pseudo conflicts with REGNO, then we consider REGNO a
2235 poor choice for a reload regno. */
2236 a = ira_regno_allocno_map[x_regno];
2237 n = ALLOCNO_NUM_OBJECTS (a);
2238 for (i = 0; i < n; i++)
2240 ira_object_t obj = ALLOCNO_OBJECT (a, i);
2241 if (TEST_HARD_REG_BIT (OBJECT_TOTAL_CONFLICT_HARD_REGS (obj), regno))
2242 return true;
2244 return false;
2247 /* Return nonzero if REGNO is a particularly bad choice for reloading
2248 IN or OUT. */
2249 bool
2250 ira_bad_reload_regno (int regno, rtx in, rtx out)
2252 return (ira_bad_reload_regno_1 (regno, in)
2253 || ira_bad_reload_regno_1 (regno, out));
2256 /* Add register clobbers from asm statements. */
2257 static void
2258 compute_regs_asm_clobbered (void)
2260 basic_block bb;
2262 FOR_EACH_BB_FN (bb, cfun)
2264 rtx_insn *insn;
2265 FOR_BB_INSNS_REVERSE (bb, insn)
2267 df_ref def;
2269 if (NONDEBUG_INSN_P (insn) && extract_asm_operands (PATTERN (insn)))
2270 FOR_EACH_INSN_DEF (def, insn)
2272 unsigned int dregno = DF_REF_REGNO (def);
2273 if (HARD_REGISTER_NUM_P (dregno))
2274 add_to_hard_reg_set (&crtl->asm_clobbers,
2275 GET_MODE (DF_REF_REAL_REG (def)),
2276 dregno);
2283 /* Set up ELIMINABLE_REGSET, IRA_NO_ALLOC_REGS, and
2284 REGS_EVER_LIVE. */
2285 void
2286 ira_setup_eliminable_regset (void)
2288 #ifdef ELIMINABLE_REGS
2289 int i;
2290 static const struct {const int from, to; } eliminables[] = ELIMINABLE_REGS;
2291 #endif
2292 /* FIXME: If EXIT_IGNORE_STACK is set, we will not save and restore
2293 sp for alloca. So we can't eliminate the frame pointer in that
2294 case. At some point, we should improve this by emitting the
2295 sp-adjusting insns for this case. */
2296 frame_pointer_needed
2297 = (! flag_omit_frame_pointer
2298 || (cfun->calls_alloca && EXIT_IGNORE_STACK)
2299 /* We need the frame pointer to catch stack overflow exceptions if
2300 the stack pointer is moving (as for the alloca case just above). */
2301 || (STACK_CHECK_MOVING_SP
2302 && flag_stack_check
2303 && flag_exceptions
2304 && cfun->can_throw_non_call_exceptions)
2305 || crtl->accesses_prior_frames
2306 || (SUPPORTS_STACK_ALIGNMENT && crtl->stack_realign_needed)
2307 /* We need a frame pointer for all Cilk Plus functions that use
2308 Cilk keywords. */
2309 || (flag_cilkplus && cfun->is_cilk_function)
2310 || targetm.frame_pointer_required ());
2312 /* The chance that FRAME_POINTER_NEEDED is changed from inspecting
2313 RTL is very small. So if we use frame pointer for RA and RTL
2314 actually prevents this, we will spill pseudos assigned to the
2315 frame pointer in LRA. */
2317 if (frame_pointer_needed)
2318 df_set_regs_ever_live (HARD_FRAME_POINTER_REGNUM, true);
2320 COPY_HARD_REG_SET (ira_no_alloc_regs, no_unit_alloc_regs);
2321 CLEAR_HARD_REG_SET (eliminable_regset);
2323 compute_regs_asm_clobbered ();
2325 /* Build the regset of all eliminable registers and show we can't
2326 use those that we already know won't be eliminated. */
2327 #ifdef ELIMINABLE_REGS
2328 for (i = 0; i < (int) ARRAY_SIZE (eliminables); i++)
2330 bool cannot_elim
2331 = (! targetm.can_eliminate (eliminables[i].from, eliminables[i].to)
2332 || (eliminables[i].to == STACK_POINTER_REGNUM && frame_pointer_needed));
2334 if (!TEST_HARD_REG_BIT (crtl->asm_clobbers, eliminables[i].from))
2336 SET_HARD_REG_BIT (eliminable_regset, eliminables[i].from);
2338 if (cannot_elim)
2339 SET_HARD_REG_BIT (ira_no_alloc_regs, eliminables[i].from);
2341 else if (cannot_elim)
2342 error ("%s cannot be used in asm here",
2343 reg_names[eliminables[i].from]);
2344 else
2345 df_set_regs_ever_live (eliminables[i].from, true);
2347 #if !HARD_FRAME_POINTER_IS_FRAME_POINTER
2348 if (!TEST_HARD_REG_BIT (crtl->asm_clobbers, HARD_FRAME_POINTER_REGNUM))
2350 SET_HARD_REG_BIT (eliminable_regset, HARD_FRAME_POINTER_REGNUM);
2351 if (frame_pointer_needed)
2352 SET_HARD_REG_BIT (ira_no_alloc_regs, HARD_FRAME_POINTER_REGNUM);
2354 else if (frame_pointer_needed)
2355 error ("%s cannot be used in asm here",
2356 reg_names[HARD_FRAME_POINTER_REGNUM]);
2357 else
2358 df_set_regs_ever_live (HARD_FRAME_POINTER_REGNUM, true);
2359 #endif
2361 #else
2362 if (!TEST_HARD_REG_BIT (crtl->asm_clobbers, HARD_FRAME_POINTER_REGNUM))
2364 SET_HARD_REG_BIT (eliminable_regset, FRAME_POINTER_REGNUM);
2365 if (frame_pointer_needed)
2366 SET_HARD_REG_BIT (ira_no_alloc_regs, FRAME_POINTER_REGNUM);
2368 else if (frame_pointer_needed)
2369 error ("%s cannot be used in asm here", reg_names[FRAME_POINTER_REGNUM]);
2370 else
2371 df_set_regs_ever_live (FRAME_POINTER_REGNUM, true);
2372 #endif
2377 /* Vector of substitutions of register numbers,
2378 used to map pseudo regs into hardware regs.
2379 This is set up as a result of register allocation.
2380 Element N is the hard reg assigned to pseudo reg N,
2381 or is -1 if no hard reg was assigned.
2382 If N is a hard reg number, element N is N. */
2383 short *reg_renumber;
2385 /* Set up REG_RENUMBER and CALLER_SAVE_NEEDED (used by reload) from
2386 the allocation found by IRA. */
2387 static void
2388 setup_reg_renumber (void)
2390 int regno, hard_regno;
2391 ira_allocno_t a;
2392 ira_allocno_iterator ai;
2394 caller_save_needed = 0;
2395 FOR_EACH_ALLOCNO (a, ai)
2397 if (ira_use_lra_p && ALLOCNO_CAP_MEMBER (a) != NULL)
2398 continue;
2399 /* There are no caps at this point. */
2400 ira_assert (ALLOCNO_CAP_MEMBER (a) == NULL);
2401 if (! ALLOCNO_ASSIGNED_P (a))
2402 /* It can happen if A is not referenced but partially anticipated
2403 somewhere in a region. */
2404 ALLOCNO_ASSIGNED_P (a) = true;
2405 ira_free_allocno_updated_costs (a);
2406 hard_regno = ALLOCNO_HARD_REGNO (a);
2407 regno = ALLOCNO_REGNO (a);
2408 reg_renumber[regno] = (hard_regno < 0 ? -1 : hard_regno);
2409 if (hard_regno >= 0)
2411 int i, nwords;
2412 enum reg_class pclass;
2413 ira_object_t obj;
2415 pclass = ira_pressure_class_translate[REGNO_REG_CLASS (hard_regno)];
2416 nwords = ALLOCNO_NUM_OBJECTS (a);
2417 for (i = 0; i < nwords; i++)
2419 obj = ALLOCNO_OBJECT (a, i);
2420 IOR_COMPL_HARD_REG_SET (OBJECT_TOTAL_CONFLICT_HARD_REGS (obj),
2421 reg_class_contents[pclass]);
2423 if (ALLOCNO_CALLS_CROSSED_NUM (a) != 0
2424 && ira_hard_reg_set_intersection_p (hard_regno, ALLOCNO_MODE (a),
2425 call_used_reg_set))
2427 ira_assert (!optimize || flag_caller_saves
2428 || (ALLOCNO_CALLS_CROSSED_NUM (a)
2429 == ALLOCNO_CHEAP_CALLS_CROSSED_NUM (a))
2430 || regno >= ira_reg_equiv_len
2431 || ira_equiv_no_lvalue_p (regno));
2432 caller_save_needed = 1;
2438 /* Set up allocno assignment flags for further allocation
2439 improvements. */
2440 static void
2441 setup_allocno_assignment_flags (void)
2443 int hard_regno;
2444 ira_allocno_t a;
2445 ira_allocno_iterator ai;
2447 FOR_EACH_ALLOCNO (a, ai)
2449 if (! ALLOCNO_ASSIGNED_P (a))
2450 /* It can happen if A is not referenced but partially anticipated
2451 somewhere in a region. */
2452 ira_free_allocno_updated_costs (a);
2453 hard_regno = ALLOCNO_HARD_REGNO (a);
2454 /* Don't assign hard registers to allocnos which are destination
2455 of removed store at the end of loop. It has no sense to keep
2456 the same value in different hard registers. It is also
2457 impossible to assign hard registers correctly to such
2458 allocnos because the cost info and info about intersected
2459 calls are incorrect for them. */
2460 ALLOCNO_ASSIGNED_P (a) = (hard_regno >= 0
2461 || ALLOCNO_EMIT_DATA (a)->mem_optimized_dest_p
2462 || (ALLOCNO_MEMORY_COST (a)
2463 - ALLOCNO_CLASS_COST (a)) < 0);
2464 ira_assert
2465 (hard_regno < 0
2466 || ira_hard_reg_in_set_p (hard_regno, ALLOCNO_MODE (a),
2467 reg_class_contents[ALLOCNO_CLASS (a)]));
2471 /* Evaluate overall allocation cost and the costs for using hard
2472 registers and memory for allocnos. */
2473 static void
2474 calculate_allocation_cost (void)
2476 int hard_regno, cost;
2477 ira_allocno_t a;
2478 ira_allocno_iterator ai;
2480 ira_overall_cost = ira_reg_cost = ira_mem_cost = 0;
2481 FOR_EACH_ALLOCNO (a, ai)
2483 hard_regno = ALLOCNO_HARD_REGNO (a);
2484 ira_assert (hard_regno < 0
2485 || (ira_hard_reg_in_set_p
2486 (hard_regno, ALLOCNO_MODE (a),
2487 reg_class_contents[ALLOCNO_CLASS (a)])));
2488 if (hard_regno < 0)
2490 cost = ALLOCNO_MEMORY_COST (a);
2491 ira_mem_cost += cost;
2493 else if (ALLOCNO_HARD_REG_COSTS (a) != NULL)
2495 cost = (ALLOCNO_HARD_REG_COSTS (a)
2496 [ira_class_hard_reg_index
2497 [ALLOCNO_CLASS (a)][hard_regno]]);
2498 ira_reg_cost += cost;
2500 else
2502 cost = ALLOCNO_CLASS_COST (a);
2503 ira_reg_cost += cost;
2505 ira_overall_cost += cost;
2508 if (internal_flag_ira_verbose > 0 && ira_dump_file != NULL)
2510 fprintf (ira_dump_file,
2511 "+++Costs: overall %" PRId64
2512 ", reg %" PRId64
2513 ", mem %" PRId64
2514 ", ld %" PRId64
2515 ", st %" PRId64
2516 ", move %" PRId64,
2517 ira_overall_cost, ira_reg_cost, ira_mem_cost,
2518 ira_load_cost, ira_store_cost, ira_shuffle_cost);
2519 fprintf (ira_dump_file, "\n+++ move loops %d, new jumps %d\n",
2520 ira_move_loops_num, ira_additional_jumps_num);
2525 #ifdef ENABLE_IRA_CHECKING
2526 /* Check the correctness of the allocation. We do need this because
2527 of complicated code to transform more one region internal
2528 representation into one region representation. */
2529 static void
2530 check_allocation (void)
2532 ira_allocno_t a;
2533 int hard_regno, nregs, conflict_nregs;
2534 ira_allocno_iterator ai;
2536 FOR_EACH_ALLOCNO (a, ai)
2538 int n = ALLOCNO_NUM_OBJECTS (a);
2539 int i;
2541 if (ALLOCNO_CAP_MEMBER (a) != NULL
2542 || (hard_regno = ALLOCNO_HARD_REGNO (a)) < 0)
2543 continue;
2544 nregs = hard_regno_nregs[hard_regno][ALLOCNO_MODE (a)];
2545 if (nregs == 1)
2546 /* We allocated a single hard register. */
2547 n = 1;
2548 else if (n > 1)
2549 /* We allocated multiple hard registers, and we will test
2550 conflicts in a granularity of single hard regs. */
2551 nregs = 1;
2553 for (i = 0; i < n; i++)
2555 ira_object_t obj = ALLOCNO_OBJECT (a, i);
2556 ira_object_t conflict_obj;
2557 ira_object_conflict_iterator oci;
2558 int this_regno = hard_regno;
2559 if (n > 1)
2561 if (REG_WORDS_BIG_ENDIAN)
2562 this_regno += n - i - 1;
2563 else
2564 this_regno += i;
2566 FOR_EACH_OBJECT_CONFLICT (obj, conflict_obj, oci)
2568 ira_allocno_t conflict_a = OBJECT_ALLOCNO (conflict_obj);
2569 int conflict_hard_regno = ALLOCNO_HARD_REGNO (conflict_a);
2570 if (conflict_hard_regno < 0)
2571 continue;
2573 conflict_nregs
2574 = (hard_regno_nregs
2575 [conflict_hard_regno][ALLOCNO_MODE (conflict_a)]);
2577 if (ALLOCNO_NUM_OBJECTS (conflict_a) > 1
2578 && conflict_nregs == ALLOCNO_NUM_OBJECTS (conflict_a))
2580 if (REG_WORDS_BIG_ENDIAN)
2581 conflict_hard_regno += (ALLOCNO_NUM_OBJECTS (conflict_a)
2582 - OBJECT_SUBWORD (conflict_obj) - 1);
2583 else
2584 conflict_hard_regno += OBJECT_SUBWORD (conflict_obj);
2585 conflict_nregs = 1;
2588 if ((conflict_hard_regno <= this_regno
2589 && this_regno < conflict_hard_regno + conflict_nregs)
2590 || (this_regno <= conflict_hard_regno
2591 && conflict_hard_regno < this_regno + nregs))
2593 fprintf (stderr, "bad allocation for %d and %d\n",
2594 ALLOCNO_REGNO (a), ALLOCNO_REGNO (conflict_a));
2595 gcc_unreachable ();
2601 #endif
2603 /* Allocate REG_EQUIV_INIT. Set up it from IRA_REG_EQUIV which should
2604 be already calculated. */
2605 static void
2606 setup_reg_equiv_init (void)
2608 int i;
2609 int max_regno = max_reg_num ();
2611 for (i = 0; i < max_regno; i++)
2612 reg_equiv_init (i) = ira_reg_equiv[i].init_insns;
2615 /* Update equiv regno from movement of FROM_REGNO to TO_REGNO. INSNS
2616 are insns which were generated for such movement. It is assumed
2617 that FROM_REGNO and TO_REGNO always have the same value at the
2618 point of any move containing such registers. This function is used
2619 to update equiv info for register shuffles on the region borders
2620 and for caller save/restore insns. */
2621 void
2622 ira_update_equiv_info_by_shuffle_insn (int to_regno, int from_regno, rtx_insn *insns)
2624 rtx_insn *insn;
2625 rtx x, note;
2627 if (! ira_reg_equiv[from_regno].defined_p
2628 && (! ira_reg_equiv[to_regno].defined_p
2629 || ((x = ira_reg_equiv[to_regno].memory) != NULL_RTX
2630 && ! MEM_READONLY_P (x))))
2631 return;
2632 insn = insns;
2633 if (NEXT_INSN (insn) != NULL_RTX)
2635 if (! ira_reg_equiv[to_regno].defined_p)
2637 ira_assert (ira_reg_equiv[to_regno].init_insns == NULL_RTX);
2638 return;
2640 ira_reg_equiv[to_regno].defined_p = false;
2641 ira_reg_equiv[to_regno].memory
2642 = ira_reg_equiv[to_regno].constant
2643 = ira_reg_equiv[to_regno].invariant
2644 = ira_reg_equiv[to_regno].init_insns = NULL;
2645 if (internal_flag_ira_verbose > 3 && ira_dump_file != NULL)
2646 fprintf (ira_dump_file,
2647 " Invalidating equiv info for reg %d\n", to_regno);
2648 return;
2650 /* It is possible that FROM_REGNO still has no equivalence because
2651 in shuffles to_regno<-from_regno and from_regno<-to_regno the 2nd
2652 insn was not processed yet. */
2653 if (ira_reg_equiv[from_regno].defined_p)
2655 ira_reg_equiv[to_regno].defined_p = true;
2656 if ((x = ira_reg_equiv[from_regno].memory) != NULL_RTX)
2658 ira_assert (ira_reg_equiv[from_regno].invariant == NULL_RTX
2659 && ira_reg_equiv[from_regno].constant == NULL_RTX);
2660 ira_assert (ira_reg_equiv[to_regno].memory == NULL_RTX
2661 || rtx_equal_p (ira_reg_equiv[to_regno].memory, x));
2662 ira_reg_equiv[to_regno].memory = x;
2663 if (! MEM_READONLY_P (x))
2664 /* We don't add the insn to insn init list because memory
2665 equivalence is just to say what memory is better to use
2666 when the pseudo is spilled. */
2667 return;
2669 else if ((x = ira_reg_equiv[from_regno].constant) != NULL_RTX)
2671 ira_assert (ira_reg_equiv[from_regno].invariant == NULL_RTX);
2672 ira_assert (ira_reg_equiv[to_regno].constant == NULL_RTX
2673 || rtx_equal_p (ira_reg_equiv[to_regno].constant, x));
2674 ira_reg_equiv[to_regno].constant = x;
2676 else
2678 x = ira_reg_equiv[from_regno].invariant;
2679 ira_assert (x != NULL_RTX);
2680 ira_assert (ira_reg_equiv[to_regno].invariant == NULL_RTX
2681 || rtx_equal_p (ira_reg_equiv[to_regno].invariant, x));
2682 ira_reg_equiv[to_regno].invariant = x;
2684 if (find_reg_note (insn, REG_EQUIV, x) == NULL_RTX)
2686 note = set_unique_reg_note (insn, REG_EQUIV, x);
2687 gcc_assert (note != NULL_RTX);
2688 if (internal_flag_ira_verbose > 3 && ira_dump_file != NULL)
2690 fprintf (ira_dump_file,
2691 " Adding equiv note to insn %u for reg %d ",
2692 INSN_UID (insn), to_regno);
2693 dump_value_slim (ira_dump_file, x, 1);
2694 fprintf (ira_dump_file, "\n");
2698 ira_reg_equiv[to_regno].init_insns
2699 = gen_rtx_INSN_LIST (VOIDmode, insn,
2700 ira_reg_equiv[to_regno].init_insns);
2701 if (internal_flag_ira_verbose > 3 && ira_dump_file != NULL)
2702 fprintf (ira_dump_file,
2703 " Adding equiv init move insn %u to reg %d\n",
2704 INSN_UID (insn), to_regno);
2707 /* Fix values of array REG_EQUIV_INIT after live range splitting done
2708 by IRA. */
2709 static void
2710 fix_reg_equiv_init (void)
2712 int max_regno = max_reg_num ();
2713 int i, new_regno, max;
2714 rtx x, prev, next, insn, set;
2716 if (max_regno_before_ira < max_regno)
2718 max = vec_safe_length (reg_equivs);
2719 grow_reg_equivs ();
2720 for (i = FIRST_PSEUDO_REGISTER; i < max; i++)
2721 for (prev = NULL_RTX, x = reg_equiv_init (i);
2722 x != NULL_RTX;
2723 x = next)
2725 next = XEXP (x, 1);
2726 insn = XEXP (x, 0);
2727 set = single_set (as_a <rtx_insn *> (insn));
2728 ira_assert (set != NULL_RTX
2729 && (REG_P (SET_DEST (set)) || REG_P (SET_SRC (set))));
2730 if (REG_P (SET_DEST (set))
2731 && ((int) REGNO (SET_DEST (set)) == i
2732 || (int) ORIGINAL_REGNO (SET_DEST (set)) == i))
2733 new_regno = REGNO (SET_DEST (set));
2734 else if (REG_P (SET_SRC (set))
2735 && ((int) REGNO (SET_SRC (set)) == i
2736 || (int) ORIGINAL_REGNO (SET_SRC (set)) == i))
2737 new_regno = REGNO (SET_SRC (set));
2738 else
2739 gcc_unreachable ();
2740 if (new_regno == i)
2741 prev = x;
2742 else
2744 /* Remove the wrong list element. */
2745 if (prev == NULL_RTX)
2746 reg_equiv_init (i) = next;
2747 else
2748 XEXP (prev, 1) = next;
2749 XEXP (x, 1) = reg_equiv_init (new_regno);
2750 reg_equiv_init (new_regno) = x;
2756 #ifdef ENABLE_IRA_CHECKING
2757 /* Print redundant memory-memory copies. */
2758 static void
2759 print_redundant_copies (void)
2761 int hard_regno;
2762 ira_allocno_t a;
2763 ira_copy_t cp, next_cp;
2764 ira_allocno_iterator ai;
2766 FOR_EACH_ALLOCNO (a, ai)
2768 if (ALLOCNO_CAP_MEMBER (a) != NULL)
2769 /* It is a cap. */
2770 continue;
2771 hard_regno = ALLOCNO_HARD_REGNO (a);
2772 if (hard_regno >= 0)
2773 continue;
2774 for (cp = ALLOCNO_COPIES (a); cp != NULL; cp = next_cp)
2775 if (cp->first == a)
2776 next_cp = cp->next_first_allocno_copy;
2777 else
2779 next_cp = cp->next_second_allocno_copy;
2780 if (internal_flag_ira_verbose > 4 && ira_dump_file != NULL
2781 && cp->insn != NULL_RTX
2782 && ALLOCNO_HARD_REGNO (cp->first) == hard_regno)
2783 fprintf (ira_dump_file,
2784 " Redundant move from %d(freq %d):%d\n",
2785 INSN_UID (cp->insn), cp->freq, hard_regno);
2789 #endif
2791 /* Setup preferred and alternative classes for new pseudo-registers
2792 created by IRA starting with START. */
2793 static void
2794 setup_preferred_alternate_classes_for_new_pseudos (int start)
2796 int i, old_regno;
2797 int max_regno = max_reg_num ();
2799 for (i = start; i < max_regno; i++)
2801 old_regno = ORIGINAL_REGNO (regno_reg_rtx[i]);
2802 ira_assert (i != old_regno);
2803 setup_reg_classes (i, reg_preferred_class (old_regno),
2804 reg_alternate_class (old_regno),
2805 reg_allocno_class (old_regno));
2806 if (internal_flag_ira_verbose > 2 && ira_dump_file != NULL)
2807 fprintf (ira_dump_file,
2808 " New r%d: setting preferred %s, alternative %s\n",
2809 i, reg_class_names[reg_preferred_class (old_regno)],
2810 reg_class_names[reg_alternate_class (old_regno)]);
2815 /* The number of entries allocated in reg_info. */
2816 static int allocated_reg_info_size;
2818 /* Regional allocation can create new pseudo-registers. This function
2819 expands some arrays for pseudo-registers. */
2820 static void
2821 expand_reg_info (void)
2823 int i;
2824 int size = max_reg_num ();
2826 resize_reg_info ();
2827 for (i = allocated_reg_info_size; i < size; i++)
2828 setup_reg_classes (i, GENERAL_REGS, ALL_REGS, GENERAL_REGS);
2829 setup_preferred_alternate_classes_for_new_pseudos (allocated_reg_info_size);
2830 allocated_reg_info_size = size;
2833 /* Return TRUE if there is too high register pressure in the function.
2834 It is used to decide when stack slot sharing is worth to do. */
2835 static bool
2836 too_high_register_pressure_p (void)
2838 int i;
2839 enum reg_class pclass;
2841 for (i = 0; i < ira_pressure_classes_num; i++)
2843 pclass = ira_pressure_classes[i];
2844 if (ira_loop_tree_root->reg_pressure[pclass] > 10000)
2845 return true;
2847 return false;
2852 /* Indicate that hard register number FROM was eliminated and replaced with
2853 an offset from hard register number TO. The status of hard registers live
2854 at the start of a basic block is updated by replacing a use of FROM with
2855 a use of TO. */
2857 void
2858 mark_elimination (int from, int to)
2860 basic_block bb;
2861 bitmap r;
2863 FOR_EACH_BB_FN (bb, cfun)
2865 r = DF_LR_IN (bb);
2866 if (bitmap_bit_p (r, from))
2868 bitmap_clear_bit (r, from);
2869 bitmap_set_bit (r, to);
2871 if (! df_live)
2872 continue;
2873 r = DF_LIVE_IN (bb);
2874 if (bitmap_bit_p (r, from))
2876 bitmap_clear_bit (r, from);
2877 bitmap_set_bit (r, to);
2884 /* The length of the following array. */
2885 int ira_reg_equiv_len;
2887 /* Info about equiv. info for each register. */
2888 struct ira_reg_equiv_s *ira_reg_equiv;
2890 /* Expand ira_reg_equiv if necessary. */
2891 void
2892 ira_expand_reg_equiv (void)
2894 int old = ira_reg_equiv_len;
2896 if (ira_reg_equiv_len > max_reg_num ())
2897 return;
2898 ira_reg_equiv_len = max_reg_num () * 3 / 2 + 1;
2899 ira_reg_equiv
2900 = (struct ira_reg_equiv_s *) xrealloc (ira_reg_equiv,
2901 ira_reg_equiv_len
2902 * sizeof (struct ira_reg_equiv_s));
2903 gcc_assert (old < ira_reg_equiv_len);
2904 memset (ira_reg_equiv + old, 0,
2905 sizeof (struct ira_reg_equiv_s) * (ira_reg_equiv_len - old));
2908 static void
2909 init_reg_equiv (void)
2911 ira_reg_equiv_len = 0;
2912 ira_reg_equiv = NULL;
2913 ira_expand_reg_equiv ();
2916 static void
2917 finish_reg_equiv (void)
2919 free (ira_reg_equiv);
2924 struct equivalence
2926 /* Set when a REG_EQUIV note is found or created. Use to
2927 keep track of what memory accesses might be created later,
2928 e.g. by reload. */
2929 rtx replacement;
2930 rtx *src_p;
2932 /* The list of each instruction which initializes this register.
2934 NULL indicates we know nothing about this register's equivalence
2935 properties.
2937 An INSN_LIST with a NULL insn indicates this pseudo is already
2938 known to not have a valid equivalence. */
2939 rtx_insn_list *init_insns;
2941 /* Loop depth is used to recognize equivalences which appear
2942 to be present within the same loop (or in an inner loop). */
2943 short loop_depth;
2944 /* Nonzero if this had a preexisting REG_EQUIV note. */
2945 unsigned char is_arg_equivalence : 1;
2946 /* Set when an attempt should be made to replace a register
2947 with the associated src_p entry. */
2948 unsigned char replace : 1;
2949 /* Set if this register has no known equivalence. */
2950 unsigned char no_equiv : 1;
2953 /* reg_equiv[N] (where N is a pseudo reg number) is the equivalence
2954 structure for that register. */
2955 static struct equivalence *reg_equiv;
2957 /* Used for communication between the following two functions: contains
2958 a MEM that we wish to ensure remains unchanged. */
2959 static rtx equiv_mem;
2961 /* Set nonzero if EQUIV_MEM is modified. */
2962 static int equiv_mem_modified;
2964 /* If EQUIV_MEM is modified by modifying DEST, indicate that it is modified.
2965 Called via note_stores. */
2966 static void
2967 validate_equiv_mem_from_store (rtx dest, const_rtx set ATTRIBUTE_UNUSED,
2968 void *data ATTRIBUTE_UNUSED)
2970 if ((REG_P (dest)
2971 && reg_overlap_mentioned_p (dest, equiv_mem))
2972 || (MEM_P (dest)
2973 && anti_dependence (equiv_mem, dest)))
2974 equiv_mem_modified = 1;
2977 /* Verify that no store between START and the death of REG invalidates
2978 MEMREF. MEMREF is invalidated by modifying a register used in MEMREF,
2979 by storing into an overlapping memory location, or with a non-const
2980 CALL_INSN.
2982 Return 1 if MEMREF remains valid. */
2983 static int
2984 validate_equiv_mem (rtx_insn *start, rtx reg, rtx memref)
2986 rtx_insn *insn;
2987 rtx note;
2989 equiv_mem = memref;
2990 equiv_mem_modified = 0;
2992 /* If the memory reference has side effects or is volatile, it isn't a
2993 valid equivalence. */
2994 if (side_effects_p (memref))
2995 return 0;
2997 for (insn = start; insn && ! equiv_mem_modified; insn = NEXT_INSN (insn))
2999 if (! INSN_P (insn))
3000 continue;
3002 if (find_reg_note (insn, REG_DEAD, reg))
3003 return 1;
3005 /* This used to ignore readonly memory and const/pure calls. The problem
3006 is the equivalent form may reference a pseudo which gets assigned a
3007 call clobbered hard reg. When we later replace REG with its
3008 equivalent form, the value in the call-clobbered reg has been
3009 changed and all hell breaks loose. */
3010 if (CALL_P (insn))
3011 return 0;
3013 note_stores (PATTERN (insn), validate_equiv_mem_from_store, NULL);
3015 /* If a register mentioned in MEMREF is modified via an
3016 auto-increment, we lose the equivalence. Do the same if one
3017 dies; although we could extend the life, it doesn't seem worth
3018 the trouble. */
3020 for (note = REG_NOTES (insn); note; note = XEXP (note, 1))
3021 if ((REG_NOTE_KIND (note) == REG_INC
3022 || REG_NOTE_KIND (note) == REG_DEAD)
3023 && REG_P (XEXP (note, 0))
3024 && reg_overlap_mentioned_p (XEXP (note, 0), memref))
3025 return 0;
3028 return 0;
3031 /* Returns zero if X is known to be invariant. */
3032 static int
3033 equiv_init_varies_p (rtx x)
3035 RTX_CODE code = GET_CODE (x);
3036 int i;
3037 const char *fmt;
3039 switch (code)
3041 case MEM:
3042 return !MEM_READONLY_P (x) || equiv_init_varies_p (XEXP (x, 0));
3044 case CONST:
3045 CASE_CONST_ANY:
3046 case SYMBOL_REF:
3047 case LABEL_REF:
3048 return 0;
3050 case REG:
3051 return reg_equiv[REGNO (x)].replace == 0 && rtx_varies_p (x, 0);
3053 case ASM_OPERANDS:
3054 if (MEM_VOLATILE_P (x))
3055 return 1;
3057 /* Fall through. */
3059 default:
3060 break;
3063 fmt = GET_RTX_FORMAT (code);
3064 for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
3065 if (fmt[i] == 'e')
3067 if (equiv_init_varies_p (XEXP (x, i)))
3068 return 1;
3070 else if (fmt[i] == 'E')
3072 int j;
3073 for (j = 0; j < XVECLEN (x, i); j++)
3074 if (equiv_init_varies_p (XVECEXP (x, i, j)))
3075 return 1;
3078 return 0;
3081 /* Returns nonzero if X (used to initialize register REGNO) is movable.
3082 X is only movable if the registers it uses have equivalent initializations
3083 which appear to be within the same loop (or in an inner loop) and movable
3084 or if they are not candidates for local_alloc and don't vary. */
3085 static int
3086 equiv_init_movable_p (rtx x, int regno)
3088 int i, j;
3089 const char *fmt;
3090 enum rtx_code code = GET_CODE (x);
3092 switch (code)
3094 case SET:
3095 return equiv_init_movable_p (SET_SRC (x), regno);
3097 case CC0:
3098 case CLOBBER:
3099 return 0;
3101 case PRE_INC:
3102 case PRE_DEC:
3103 case POST_INC:
3104 case POST_DEC:
3105 case PRE_MODIFY:
3106 case POST_MODIFY:
3107 return 0;
3109 case REG:
3110 return ((reg_equiv[REGNO (x)].loop_depth >= reg_equiv[regno].loop_depth
3111 && reg_equiv[REGNO (x)].replace)
3112 || (REG_BASIC_BLOCK (REGNO (x)) < NUM_FIXED_BLOCKS
3113 && ! rtx_varies_p (x, 0)));
3115 case UNSPEC_VOLATILE:
3116 return 0;
3118 case ASM_OPERANDS:
3119 if (MEM_VOLATILE_P (x))
3120 return 0;
3122 /* Fall through. */
3124 default:
3125 break;
3128 fmt = GET_RTX_FORMAT (code);
3129 for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
3130 switch (fmt[i])
3132 case 'e':
3133 if (! equiv_init_movable_p (XEXP (x, i), regno))
3134 return 0;
3135 break;
3136 case 'E':
3137 for (j = XVECLEN (x, i) - 1; j >= 0; j--)
3138 if (! equiv_init_movable_p (XVECEXP (x, i, j), regno))
3139 return 0;
3140 break;
3143 return 1;
3146 /* TRUE if X uses any registers for which reg_equiv[REGNO].replace is
3147 true. */
3148 static int
3149 contains_replace_regs (rtx x)
3151 int i, j;
3152 const char *fmt;
3153 enum rtx_code code = GET_CODE (x);
3155 switch (code)
3157 case CONST:
3158 case LABEL_REF:
3159 case SYMBOL_REF:
3160 CASE_CONST_ANY:
3161 case PC:
3162 case CC0:
3163 case HIGH:
3164 return 0;
3166 case REG:
3167 return reg_equiv[REGNO (x)].replace;
3169 default:
3170 break;
3173 fmt = GET_RTX_FORMAT (code);
3174 for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
3175 switch (fmt[i])
3177 case 'e':
3178 if (contains_replace_regs (XEXP (x, i)))
3179 return 1;
3180 break;
3181 case 'E':
3182 for (j = XVECLEN (x, i) - 1; j >= 0; j--)
3183 if (contains_replace_regs (XVECEXP (x, i, j)))
3184 return 1;
3185 break;
3188 return 0;
3191 /* TRUE if X references a memory location that would be affected by a store
3192 to MEMREF. */
3193 static int
3194 memref_referenced_p (rtx memref, rtx x)
3196 int i, j;
3197 const char *fmt;
3198 enum rtx_code code = GET_CODE (x);
3200 switch (code)
3202 case CONST:
3203 case LABEL_REF:
3204 case SYMBOL_REF:
3205 CASE_CONST_ANY:
3206 case PC:
3207 case CC0:
3208 case HIGH:
3209 case LO_SUM:
3210 return 0;
3212 case REG:
3213 return (reg_equiv[REGNO (x)].replacement
3214 && memref_referenced_p (memref,
3215 reg_equiv[REGNO (x)].replacement));
3217 case MEM:
3218 if (true_dependence (memref, VOIDmode, x))
3219 return 1;
3220 break;
3222 case SET:
3223 /* If we are setting a MEM, it doesn't count (its address does), but any
3224 other SET_DEST that has a MEM in it is referencing the MEM. */
3225 if (MEM_P (SET_DEST (x)))
3227 if (memref_referenced_p (memref, XEXP (SET_DEST (x), 0)))
3228 return 1;
3230 else if (memref_referenced_p (memref, SET_DEST (x)))
3231 return 1;
3233 return memref_referenced_p (memref, SET_SRC (x));
3235 default:
3236 break;
3239 fmt = GET_RTX_FORMAT (code);
3240 for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
3241 switch (fmt[i])
3243 case 'e':
3244 if (memref_referenced_p (memref, XEXP (x, i)))
3245 return 1;
3246 break;
3247 case 'E':
3248 for (j = XVECLEN (x, i) - 1; j >= 0; j--)
3249 if (memref_referenced_p (memref, XVECEXP (x, i, j)))
3250 return 1;
3251 break;
3254 return 0;
3257 /* TRUE if some insn in the range (START, END] references a memory location
3258 that would be affected by a store to MEMREF. */
3259 static int
3260 memref_used_between_p (rtx memref, rtx_insn *start, rtx_insn *end)
3262 rtx_insn *insn;
3264 for (insn = NEXT_INSN (start); insn != NEXT_INSN (end);
3265 insn = NEXT_INSN (insn))
3267 if (!NONDEBUG_INSN_P (insn))
3268 continue;
3270 if (memref_referenced_p (memref, PATTERN (insn)))
3271 return 1;
3273 /* Nonconst functions may access memory. */
3274 if (CALL_P (insn) && (! RTL_CONST_CALL_P (insn)))
3275 return 1;
3278 return 0;
3281 /* Mark REG as having no known equivalence.
3282 Some instructions might have been processed before and furnished
3283 with REG_EQUIV notes for this register; these notes will have to be
3284 removed.
3285 STORE is the piece of RTL that does the non-constant / conflicting
3286 assignment - a SET, CLOBBER or REG_INC note. It is currently not used,
3287 but needs to be there because this function is called from note_stores. */
3288 static void
3289 no_equiv (rtx reg, const_rtx store ATTRIBUTE_UNUSED,
3290 void *data ATTRIBUTE_UNUSED)
3292 int regno;
3293 rtx_insn_list *list;
3295 if (!REG_P (reg))
3296 return;
3297 regno = REGNO (reg);
3298 reg_equiv[regno].no_equiv = 1;
3299 list = reg_equiv[regno].init_insns;
3300 if (list && list->insn () == NULL)
3301 return;
3302 reg_equiv[regno].init_insns = gen_rtx_INSN_LIST (VOIDmode, NULL_RTX, NULL);
3303 reg_equiv[regno].replacement = NULL_RTX;
3304 /* This doesn't matter for equivalences made for argument registers, we
3305 should keep their initialization insns. */
3306 if (reg_equiv[regno].is_arg_equivalence)
3307 return;
3308 ira_reg_equiv[regno].defined_p = false;
3309 ira_reg_equiv[regno].init_insns = NULL;
3310 for (; list; list = list->next ())
3312 rtx_insn *insn = list->insn ();
3313 remove_note (insn, find_reg_note (insn, REG_EQUIV, NULL_RTX));
3317 /* Check whether the SUBREG is a paradoxical subreg and set the result
3318 in PDX_SUBREGS. */
3320 static void
3321 set_paradoxical_subreg (rtx_insn *insn, bool *pdx_subregs)
3323 subrtx_iterator::array_type array;
3324 FOR_EACH_SUBRTX (iter, array, PATTERN (insn), NONCONST)
3326 const_rtx subreg = *iter;
3327 if (GET_CODE (subreg) == SUBREG)
3329 const_rtx reg = SUBREG_REG (subreg);
3330 if (REG_P (reg) && paradoxical_subreg_p (subreg))
3331 pdx_subregs[REGNO (reg)] = true;
3336 /* In DEBUG_INSN location adjust REGs from CLEARED_REGS bitmap to the
3337 equivalent replacement. */
3339 static rtx
3340 adjust_cleared_regs (rtx loc, const_rtx old_rtx ATTRIBUTE_UNUSED, void *data)
3342 if (REG_P (loc))
3344 bitmap cleared_regs = (bitmap) data;
3345 if (bitmap_bit_p (cleared_regs, REGNO (loc)))
3346 return simplify_replace_fn_rtx (copy_rtx (*reg_equiv[REGNO (loc)].src_p),
3347 NULL_RTX, adjust_cleared_regs, data);
3349 return NULL_RTX;
3352 /* Find registers that are equivalent to a single value throughout the
3353 compilation (either because they can be referenced in memory or are
3354 set once from a single constant). Lower their priority for a
3355 register.
3357 If such a register is only referenced once, try substituting its
3358 value into the using insn. If it succeeds, we can eliminate the
3359 register completely.
3361 Initialize init_insns in ira_reg_equiv array. */
3362 static void
3363 update_equiv_regs (void)
3365 rtx_insn *insn;
3366 basic_block bb;
3367 int loop_depth;
3368 bitmap cleared_regs;
3369 bool *pdx_subregs;
3371 /* Use pdx_subregs to show whether a reg is used in a paradoxical
3372 subreg. */
3373 pdx_subregs = XCNEWVEC (bool, max_regno);
3375 reg_equiv = XCNEWVEC (struct equivalence, max_regno);
3376 grow_reg_equivs ();
3378 init_alias_analysis ();
3380 /* Scan insns and set pdx_subregs[regno] if the reg is used in a
3381 paradoxical subreg. Don't set such reg equivalent to a mem,
3382 because lra will not substitute such equiv memory in order to
3383 prevent access beyond allocated memory for paradoxical memory subreg. */
3384 FOR_EACH_BB_FN (bb, cfun)
3385 FOR_BB_INSNS (bb, insn)
3386 if (NONDEBUG_INSN_P (insn))
3387 set_paradoxical_subreg (insn, pdx_subregs);
3389 /* Scan the insns and find which registers have equivalences. Do this
3390 in a separate scan of the insns because (due to -fcse-follow-jumps)
3391 a register can be set below its use. */
3392 FOR_EACH_BB_FN (bb, cfun)
3394 loop_depth = bb_loop_depth (bb);
3396 for (insn = BB_HEAD (bb);
3397 insn != NEXT_INSN (BB_END (bb));
3398 insn = NEXT_INSN (insn))
3400 rtx note;
3401 rtx set;
3402 rtx dest, src;
3403 int regno;
3405 if (! INSN_P (insn))
3406 continue;
3408 for (note = REG_NOTES (insn); note; note = XEXP (note, 1))
3409 if (REG_NOTE_KIND (note) == REG_INC)
3410 no_equiv (XEXP (note, 0), note, NULL);
3412 set = single_set (insn);
3414 /* If this insn contains more (or less) than a single SET,
3415 only mark all destinations as having no known equivalence. */
3416 if (set == NULL_RTX
3417 || side_effects_p (SET_SRC (set)))
3419 note_stores (PATTERN (insn), no_equiv, NULL);
3420 continue;
3422 else if (GET_CODE (PATTERN (insn)) == PARALLEL)
3424 int i;
3426 for (i = XVECLEN (PATTERN (insn), 0) - 1; i >= 0; i--)
3428 rtx part = XVECEXP (PATTERN (insn), 0, i);
3429 if (part != set)
3430 note_stores (part, no_equiv, NULL);
3434 dest = SET_DEST (set);
3435 src = SET_SRC (set);
3437 /* See if this is setting up the equivalence between an argument
3438 register and its stack slot. */
3439 note = find_reg_note (insn, REG_EQUIV, NULL_RTX);
3440 if (note)
3442 gcc_assert (REG_P (dest));
3443 regno = REGNO (dest);
3445 /* Note that we don't want to clear init_insns in
3446 ira_reg_equiv even if there are multiple sets of this
3447 register. */
3448 reg_equiv[regno].is_arg_equivalence = 1;
3450 /* The insn result can have equivalence memory although
3451 the equivalence is not set up by the insn. We add
3452 this insn to init insns as it is a flag for now that
3453 regno has an equivalence. We will remove the insn
3454 from init insn list later. */
3455 if (rtx_equal_p (src, XEXP (note, 0)) || MEM_P (XEXP (note, 0)))
3456 ira_reg_equiv[regno].init_insns
3457 = gen_rtx_INSN_LIST (VOIDmode, insn,
3458 ira_reg_equiv[regno].init_insns);
3460 /* Continue normally in case this is a candidate for
3461 replacements. */
3464 if (!optimize)
3465 continue;
3467 /* We only handle the case of a pseudo register being set
3468 once, or always to the same value. */
3469 /* ??? The mn10200 port breaks if we add equivalences for
3470 values that need an ADDRESS_REGS register and set them equivalent
3471 to a MEM of a pseudo. The actual problem is in the over-conservative
3472 handling of INPADDR_ADDRESS / INPUT_ADDRESS / INPUT triples in
3473 calculate_needs, but we traditionally work around this problem
3474 here by rejecting equivalences when the destination is in a register
3475 that's likely spilled. This is fragile, of course, since the
3476 preferred class of a pseudo depends on all instructions that set
3477 or use it. */
3479 if (!REG_P (dest)
3480 || (regno = REGNO (dest)) < FIRST_PSEUDO_REGISTER
3481 || (reg_equiv[regno].init_insns
3482 && reg_equiv[regno].init_insns->insn () == NULL)
3483 || (targetm.class_likely_spilled_p (reg_preferred_class (regno))
3484 && MEM_P (src) && ! reg_equiv[regno].is_arg_equivalence))
3486 /* This might be setting a SUBREG of a pseudo, a pseudo that is
3487 also set somewhere else to a constant. */
3488 note_stores (set, no_equiv, NULL);
3489 continue;
3492 /* Don't set reg (if pdx_subregs[regno] == true) equivalent to a mem. */
3493 if (MEM_P (src) && pdx_subregs[regno])
3495 note_stores (set, no_equiv, NULL);
3496 continue;
3499 note = find_reg_note (insn, REG_EQUAL, NULL_RTX);
3501 /* cse sometimes generates function invariants, but doesn't put a
3502 REG_EQUAL note on the insn. Since this note would be redundant,
3503 there's no point creating it earlier than here. */
3504 if (! note && ! rtx_varies_p (src, 0))
3505 note = set_unique_reg_note (insn, REG_EQUAL, copy_rtx (src));
3507 /* Don't bother considering a REG_EQUAL note containing an EXPR_LIST
3508 since it represents a function call. */
3509 if (note && GET_CODE (XEXP (note, 0)) == EXPR_LIST)
3510 note = NULL_RTX;
3512 if (DF_REG_DEF_COUNT (regno) != 1)
3514 bool equal_p = true;
3515 rtx_insn_list *list;
3517 /* If we have already processed this pseudo and determined it
3518 can not have an equivalence, then honor that decision. */
3519 if (reg_equiv[regno].no_equiv)
3520 continue;
3522 if (! note
3523 || rtx_varies_p (XEXP (note, 0), 0)
3524 || (reg_equiv[regno].replacement
3525 && ! rtx_equal_p (XEXP (note, 0),
3526 reg_equiv[regno].replacement)))
3528 no_equiv (dest, set, NULL);
3529 continue;
3532 list = reg_equiv[regno].init_insns;
3533 for (; list; list = list->next ())
3535 rtx note_tmp;
3536 rtx_insn *insn_tmp;
3538 insn_tmp = list->insn ();
3539 note_tmp = find_reg_note (insn_tmp, REG_EQUAL, NULL_RTX);
3540 gcc_assert (note_tmp);
3541 if (! rtx_equal_p (XEXP (note, 0), XEXP (note_tmp, 0)))
3543 equal_p = false;
3544 break;
3548 if (! equal_p)
3550 no_equiv (dest, set, NULL);
3551 continue;
3555 /* Record this insn as initializing this register. */
3556 reg_equiv[regno].init_insns
3557 = gen_rtx_INSN_LIST (VOIDmode, insn, reg_equiv[regno].init_insns);
3559 /* If this register is known to be equal to a constant, record that
3560 it is always equivalent to the constant. */
3561 if (DF_REG_DEF_COUNT (regno) == 1
3562 && note && ! rtx_varies_p (XEXP (note, 0), 0))
3564 rtx note_value = XEXP (note, 0);
3565 remove_note (insn, note);
3566 set_unique_reg_note (insn, REG_EQUIV, note_value);
3569 /* If this insn introduces a "constant" register, decrease the priority
3570 of that register. Record this insn if the register is only used once
3571 more and the equivalence value is the same as our source.
3573 The latter condition is checked for two reasons: First, it is an
3574 indication that it may be more efficient to actually emit the insn
3575 as written (if no registers are available, reload will substitute
3576 the equivalence). Secondly, it avoids problems with any registers
3577 dying in this insn whose death notes would be missed.
3579 If we don't have a REG_EQUIV note, see if this insn is loading
3580 a register used only in one basic block from a MEM. If so, and the
3581 MEM remains unchanged for the life of the register, add a REG_EQUIV
3582 note. */
3583 note = find_reg_note (insn, REG_EQUIV, NULL_RTX);
3585 if (note == NULL_RTX && REG_BASIC_BLOCK (regno) >= NUM_FIXED_BLOCKS
3586 && MEM_P (SET_SRC (set))
3587 && validate_equiv_mem (insn, dest, SET_SRC (set)))
3588 note = set_unique_reg_note (insn, REG_EQUIV, copy_rtx (SET_SRC (set)));
3590 if (note)
3592 int regno = REGNO (dest);
3593 rtx x = XEXP (note, 0);
3595 /* If we haven't done so, record for reload that this is an
3596 equivalencing insn. */
3597 if (!reg_equiv[regno].is_arg_equivalence)
3598 ira_reg_equiv[regno].init_insns
3599 = gen_rtx_INSN_LIST (VOIDmode, insn,
3600 ira_reg_equiv[regno].init_insns);
3602 reg_equiv[regno].replacement = x;
3603 reg_equiv[regno].src_p = &SET_SRC (set);
3604 reg_equiv[regno].loop_depth = (short) loop_depth;
3606 /* Don't mess with things live during setjmp. */
3607 if (REG_LIVE_LENGTH (regno) >= 0 && optimize)
3609 /* Note that the statement below does not affect the priority
3610 in local-alloc! */
3611 REG_LIVE_LENGTH (regno) *= 2;
3613 /* If the register is referenced exactly twice, meaning it is
3614 set once and used once, indicate that the reference may be
3615 replaced by the equivalence we computed above. Do this
3616 even if the register is only used in one block so that
3617 dependencies can be handled where the last register is
3618 used in a different block (i.e. HIGH / LO_SUM sequences)
3619 and to reduce the number of registers alive across
3620 calls. */
3622 if (REG_N_REFS (regno) == 2
3623 && (rtx_equal_p (x, src)
3624 || ! equiv_init_varies_p (src))
3625 && NONJUMP_INSN_P (insn)
3626 && equiv_init_movable_p (PATTERN (insn), regno))
3627 reg_equiv[regno].replace = 1;
3633 if (!optimize)
3634 goto out;
3636 /* A second pass, to gather additional equivalences with memory. This needs
3637 to be done after we know which registers we are going to replace. */
3639 for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
3641 rtx set, src, dest;
3642 unsigned regno;
3644 if (! INSN_P (insn))
3645 continue;
3647 set = single_set (insn);
3648 if (! set)
3649 continue;
3651 dest = SET_DEST (set);
3652 src = SET_SRC (set);
3654 /* If this sets a MEM to the contents of a REG that is only used
3655 in a single basic block, see if the register is always equivalent
3656 to that memory location and if moving the store from INSN to the
3657 insn that set REG is safe. If so, put a REG_EQUIV note on the
3658 initializing insn.
3660 Don't add a REG_EQUIV note if the insn already has one. The existing
3661 REG_EQUIV is likely more useful than the one we are adding.
3663 If one of the regs in the address has reg_equiv[REGNO].replace set,
3664 then we can't add this REG_EQUIV note. The reg_equiv[REGNO].replace
3665 optimization may move the set of this register immediately before
3666 insn, which puts it after reg_equiv[REGNO].init_insns, and hence
3667 the mention in the REG_EQUIV note would be to an uninitialized
3668 pseudo. */
3670 if (MEM_P (dest) && REG_P (src)
3671 && (regno = REGNO (src)) >= FIRST_PSEUDO_REGISTER
3672 && REG_BASIC_BLOCK (regno) >= NUM_FIXED_BLOCKS
3673 && DF_REG_DEF_COUNT (regno) == 1
3674 && reg_equiv[regno].init_insns != NULL
3675 && reg_equiv[regno].init_insns->insn () != NULL
3676 && ! find_reg_note (XEXP (reg_equiv[regno].init_insns, 0),
3677 REG_EQUIV, NULL_RTX)
3678 && ! contains_replace_regs (XEXP (dest, 0))
3679 && ! pdx_subregs[regno])
3681 rtx_insn *init_insn =
3682 as_a <rtx_insn *> (XEXP (reg_equiv[regno].init_insns, 0));
3683 if (validate_equiv_mem (init_insn, src, dest)
3684 && ! memref_used_between_p (dest, init_insn, insn)
3685 /* Attaching a REG_EQUIV note will fail if INIT_INSN has
3686 multiple sets. */
3687 && set_unique_reg_note (init_insn, REG_EQUIV, copy_rtx (dest)))
3689 /* This insn makes the equivalence, not the one initializing
3690 the register. */
3691 ira_reg_equiv[regno].init_insns
3692 = gen_rtx_INSN_LIST (VOIDmode, insn, NULL_RTX);
3693 df_notes_rescan (init_insn);
3698 cleared_regs = BITMAP_ALLOC (NULL);
3699 /* Now scan all regs killed in an insn to see if any of them are
3700 registers only used that once. If so, see if we can replace the
3701 reference with the equivalent form. If we can, delete the
3702 initializing reference and this register will go away. If we
3703 can't replace the reference, and the initializing reference is
3704 within the same loop (or in an inner loop), then move the register
3705 initialization just before the use, so that they are in the same
3706 basic block. */
3707 FOR_EACH_BB_REVERSE_FN (bb, cfun)
3709 loop_depth = bb_loop_depth (bb);
3710 for (insn = BB_END (bb);
3711 insn != PREV_INSN (BB_HEAD (bb));
3712 insn = PREV_INSN (insn))
3714 rtx link;
3716 if (! INSN_P (insn))
3717 continue;
3719 /* Don't substitute into jumps. indirect_jump_optimize does
3720 this for anything we are prepared to handle. */
3721 if (JUMP_P (insn))
3722 continue;
3724 for (link = REG_NOTES (insn); link; link = XEXP (link, 1))
3726 if (REG_NOTE_KIND (link) == REG_DEAD
3727 /* Make sure this insn still refers to the register. */
3728 && reg_mentioned_p (XEXP (link, 0), PATTERN (insn)))
3730 int regno = REGNO (XEXP (link, 0));
3731 rtx equiv_insn;
3733 if (! reg_equiv[regno].replace
3734 || reg_equiv[regno].loop_depth < (short) loop_depth
3735 /* There is no sense to move insns if live range
3736 shrinkage or register pressure-sensitive
3737 scheduling were done because it will not
3738 improve allocation but worsen insn schedule
3739 with a big probability. */
3740 || flag_live_range_shrinkage
3741 || (flag_sched_pressure && flag_schedule_insns))
3742 continue;
3744 /* reg_equiv[REGNO].replace gets set only when
3745 REG_N_REFS[REGNO] is 2, i.e. the register is set
3746 once and used once. (If it were only set, but
3747 not used, flow would have deleted the setting
3748 insns.) Hence there can only be one insn in
3749 reg_equiv[REGNO].init_insns. */
3750 gcc_assert (reg_equiv[regno].init_insns
3751 && !XEXP (reg_equiv[regno].init_insns, 1));
3752 equiv_insn = XEXP (reg_equiv[regno].init_insns, 0);
3754 /* We may not move instructions that can throw, since
3755 that changes basic block boundaries and we are not
3756 prepared to adjust the CFG to match. */
3757 if (can_throw_internal (equiv_insn))
3758 continue;
3760 if (asm_noperands (PATTERN (equiv_insn)) < 0
3761 && validate_replace_rtx (regno_reg_rtx[regno],
3762 *(reg_equiv[regno].src_p), insn))
3764 rtx equiv_link;
3765 rtx last_link;
3766 rtx note;
3768 /* Find the last note. */
3769 for (last_link = link; XEXP (last_link, 1);
3770 last_link = XEXP (last_link, 1))
3773 /* Append the REG_DEAD notes from equiv_insn. */
3774 equiv_link = REG_NOTES (equiv_insn);
3775 while (equiv_link)
3777 note = equiv_link;
3778 equiv_link = XEXP (equiv_link, 1);
3779 if (REG_NOTE_KIND (note) == REG_DEAD)
3781 remove_note (equiv_insn, note);
3782 XEXP (last_link, 1) = note;
3783 XEXP (note, 1) = NULL_RTX;
3784 last_link = note;
3788 remove_death (regno, insn);
3789 SET_REG_N_REFS (regno, 0);
3790 REG_FREQ (regno) = 0;
3791 delete_insn (equiv_insn);
3793 reg_equiv[regno].init_insns
3794 = reg_equiv[regno].init_insns->next ();
3796 ira_reg_equiv[regno].init_insns = NULL;
3797 bitmap_set_bit (cleared_regs, regno);
3799 /* Move the initialization of the register to just before
3800 INSN. Update the flow information. */
3801 else if (prev_nondebug_insn (insn) != equiv_insn)
3803 rtx_insn *new_insn;
3805 new_insn = emit_insn_before (PATTERN (equiv_insn), insn);
3806 REG_NOTES (new_insn) = REG_NOTES (equiv_insn);
3807 REG_NOTES (equiv_insn) = 0;
3808 /* Rescan it to process the notes. */
3809 df_insn_rescan (new_insn);
3811 /* Make sure this insn is recognized before
3812 reload begins, otherwise
3813 eliminate_regs_in_insn will die. */
3814 INSN_CODE (new_insn) = INSN_CODE (equiv_insn);
3816 delete_insn (equiv_insn);
3818 XEXP (reg_equiv[regno].init_insns, 0) = new_insn;
3820 REG_BASIC_BLOCK (regno) = bb->index;
3821 REG_N_CALLS_CROSSED (regno) = 0;
3822 REG_FREQ_CALLS_CROSSED (regno) = 0;
3823 REG_N_THROWING_CALLS_CROSSED (regno) = 0;
3824 REG_LIVE_LENGTH (regno) = 2;
3826 if (insn == BB_HEAD (bb))
3827 BB_HEAD (bb) = PREV_INSN (insn);
3829 ira_reg_equiv[regno].init_insns
3830 = gen_rtx_INSN_LIST (VOIDmode, new_insn, NULL_RTX);
3831 bitmap_set_bit (cleared_regs, regno);
3838 if (!bitmap_empty_p (cleared_regs))
3840 FOR_EACH_BB_FN (bb, cfun)
3842 bitmap_and_compl_into (DF_LR_IN (bb), cleared_regs);
3843 bitmap_and_compl_into (DF_LR_OUT (bb), cleared_regs);
3844 if (! df_live)
3845 continue;
3846 bitmap_and_compl_into (DF_LIVE_IN (bb), cleared_regs);
3847 bitmap_and_compl_into (DF_LIVE_OUT (bb), cleared_regs);
3850 /* Last pass - adjust debug insns referencing cleared regs. */
3851 if (MAY_HAVE_DEBUG_INSNS)
3852 for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
3853 if (DEBUG_INSN_P (insn))
3855 rtx old_loc = INSN_VAR_LOCATION_LOC (insn);
3856 INSN_VAR_LOCATION_LOC (insn)
3857 = simplify_replace_fn_rtx (old_loc, NULL_RTX,
3858 adjust_cleared_regs,
3859 (void *) cleared_regs);
3860 if (old_loc != INSN_VAR_LOCATION_LOC (insn))
3861 df_insn_rescan (insn);
3865 BITMAP_FREE (cleared_regs);
3867 out:
3868 /* Clean up. */
3870 end_alias_analysis ();
3871 free (reg_equiv);
3872 free (pdx_subregs);
3875 /* A pass over indirect jumps, converting simple cases to direct jumps.
3876 Combine does this optimization too, but only within a basic block. */
3877 static void
3878 indirect_jump_optimize (void)
3880 basic_block bb;
3881 bool rebuild_p = false;
3883 FOR_EACH_BB_REVERSE_FN (bb, cfun)
3885 rtx_insn *insn = BB_END (bb);
3886 if (!JUMP_P (insn)
3887 || find_reg_note (insn, REG_NON_LOCAL_GOTO, NULL_RTX))
3888 continue;
3890 rtx x = pc_set (insn);
3891 if (!x || !REG_P (SET_SRC (x)))
3892 continue;
3894 int regno = REGNO (SET_SRC (x));
3895 if (DF_REG_DEF_COUNT (regno) == 1)
3897 df_ref def = DF_REG_DEF_CHAIN (regno);
3898 if (!DF_REF_IS_ARTIFICIAL (def))
3900 rtx_insn *def_insn = DF_REF_INSN (def);
3901 rtx lab = NULL_RTX;
3902 rtx set = single_set (def_insn);
3903 if (set && GET_CODE (SET_SRC (set)) == LABEL_REF)
3904 lab = SET_SRC (set);
3905 else
3907 rtx eqnote = find_reg_note (def_insn, REG_EQUAL, NULL_RTX);
3908 if (eqnote && GET_CODE (XEXP (eqnote, 0)) == LABEL_REF)
3909 lab = XEXP (eqnote, 0);
3911 if (lab && validate_replace_rtx (SET_SRC (x), lab, insn))
3912 rebuild_p = true;
3917 if (rebuild_p)
3919 timevar_push (TV_JUMP);
3920 rebuild_jump_labels (get_insns ());
3921 if (purge_all_dead_edges ())
3922 delete_unreachable_blocks ();
3923 timevar_pop (TV_JUMP);
3927 /* Set up fields memory, constant, and invariant from init_insns in
3928 the structures of array ira_reg_equiv. */
3929 static void
3930 setup_reg_equiv (void)
3932 int i;
3933 rtx_insn_list *elem, *prev_elem, *next_elem;
3934 rtx_insn *insn;
3935 rtx set, x;
3937 for (i = FIRST_PSEUDO_REGISTER; i < ira_reg_equiv_len; i++)
3938 for (prev_elem = NULL, elem = ira_reg_equiv[i].init_insns;
3939 elem;
3940 prev_elem = elem, elem = next_elem)
3942 next_elem = elem->next ();
3943 insn = elem->insn ();
3944 set = single_set (insn);
3946 /* Init insns can set up equivalence when the reg is a destination or
3947 a source (in this case the destination is memory). */
3948 if (set != 0 && (REG_P (SET_DEST (set)) || REG_P (SET_SRC (set))))
3950 if ((x = find_reg_note (insn, REG_EQUIV, NULL_RTX)) != NULL)
3952 x = XEXP (x, 0);
3953 if (REG_P (SET_DEST (set))
3954 && REGNO (SET_DEST (set)) == (unsigned int) i
3955 && ! rtx_equal_p (SET_SRC (set), x) && MEM_P (x))
3957 /* This insn reporting the equivalence but
3958 actually not setting it. Remove it from the
3959 list. */
3960 if (prev_elem == NULL)
3961 ira_reg_equiv[i].init_insns = next_elem;
3962 else
3963 XEXP (prev_elem, 1) = next_elem;
3964 elem = prev_elem;
3967 else if (REG_P (SET_DEST (set))
3968 && REGNO (SET_DEST (set)) == (unsigned int) i)
3969 x = SET_SRC (set);
3970 else
3972 gcc_assert (REG_P (SET_SRC (set))
3973 && REGNO (SET_SRC (set)) == (unsigned int) i);
3974 x = SET_DEST (set);
3976 if (! function_invariant_p (x)
3977 || ! flag_pic
3978 /* A function invariant is often CONSTANT_P but may
3979 include a register. We promise to only pass
3980 CONSTANT_P objects to LEGITIMATE_PIC_OPERAND_P. */
3981 || (CONSTANT_P (x) && LEGITIMATE_PIC_OPERAND_P (x)))
3983 /* It can happen that a REG_EQUIV note contains a MEM
3984 that is not a legitimate memory operand. As later
3985 stages of reload assume that all addresses found in
3986 the lra_regno_equiv_* arrays were originally
3987 legitimate, we ignore such REG_EQUIV notes. */
3988 if (memory_operand (x, VOIDmode))
3990 ira_reg_equiv[i].defined_p = true;
3991 ira_reg_equiv[i].memory = x;
3992 continue;
3994 else if (function_invariant_p (x))
3996 machine_mode mode;
3998 mode = GET_MODE (SET_DEST (set));
3999 if (GET_CODE (x) == PLUS
4000 || x == frame_pointer_rtx || x == arg_pointer_rtx)
4001 /* This is PLUS of frame pointer and a constant,
4002 or fp, or argp. */
4003 ira_reg_equiv[i].invariant = x;
4004 else if (targetm.legitimate_constant_p (mode, x))
4005 ira_reg_equiv[i].constant = x;
4006 else
4008 ira_reg_equiv[i].memory = force_const_mem (mode, x);
4009 if (ira_reg_equiv[i].memory == NULL_RTX)
4011 ira_reg_equiv[i].defined_p = false;
4012 ira_reg_equiv[i].init_insns = NULL;
4013 break;
4016 ira_reg_equiv[i].defined_p = true;
4017 continue;
4021 ira_reg_equiv[i].defined_p = false;
4022 ira_reg_equiv[i].init_insns = NULL;
4023 break;
4029 /* Print chain C to FILE. */
4030 static void
4031 print_insn_chain (FILE *file, struct insn_chain *c)
4033 fprintf (file, "insn=%d, ", INSN_UID (c->insn));
4034 bitmap_print (file, &c->live_throughout, "live_throughout: ", ", ");
4035 bitmap_print (file, &c->dead_or_set, "dead_or_set: ", "\n");
4039 /* Print all reload_insn_chains to FILE. */
4040 static void
4041 print_insn_chains (FILE *file)
4043 struct insn_chain *c;
4044 for (c = reload_insn_chain; c ; c = c->next)
4045 print_insn_chain (file, c);
4048 /* Return true if pseudo REGNO should be added to set live_throughout
4049 or dead_or_set of the insn chains for reload consideration. */
4050 static bool
4051 pseudo_for_reload_consideration_p (int regno)
4053 /* Consider spilled pseudos too for IRA because they still have a
4054 chance to get hard-registers in the reload when IRA is used. */
4055 return (reg_renumber[regno] >= 0 || ira_conflicts_p);
4058 /* Init LIVE_SUBREGS[ALLOCNUM] and LIVE_SUBREGS_USED[ALLOCNUM] using
4059 REG to the number of nregs, and INIT_VALUE to get the
4060 initialization. ALLOCNUM need not be the regno of REG. */
4061 static void
4062 init_live_subregs (bool init_value, sbitmap *live_subregs,
4063 bitmap live_subregs_used, int allocnum, rtx reg)
4065 unsigned int regno = REGNO (SUBREG_REG (reg));
4066 int size = GET_MODE_SIZE (GET_MODE (regno_reg_rtx[regno]));
4068 gcc_assert (size > 0);
4070 /* Been there, done that. */
4071 if (bitmap_bit_p (live_subregs_used, allocnum))
4072 return;
4074 /* Create a new one. */
4075 if (live_subregs[allocnum] == NULL)
4076 live_subregs[allocnum] = sbitmap_alloc (size);
4078 /* If the entire reg was live before blasting into subregs, we need
4079 to init all of the subregs to ones else init to 0. */
4080 if (init_value)
4081 bitmap_ones (live_subregs[allocnum]);
4082 else
4083 bitmap_clear (live_subregs[allocnum]);
4085 bitmap_set_bit (live_subregs_used, allocnum);
4088 /* Walk the insns of the current function and build reload_insn_chain,
4089 and record register life information. */
4090 static void
4091 build_insn_chain (void)
4093 unsigned int i;
4094 struct insn_chain **p = &reload_insn_chain;
4095 basic_block bb;
4096 struct insn_chain *c = NULL;
4097 struct insn_chain *next = NULL;
4098 bitmap live_relevant_regs = BITMAP_ALLOC (NULL);
4099 bitmap elim_regset = BITMAP_ALLOC (NULL);
4100 /* live_subregs is a vector used to keep accurate information about
4101 which hardregs are live in multiword pseudos. live_subregs and
4102 live_subregs_used are indexed by pseudo number. The live_subreg
4103 entry for a particular pseudo is only used if the corresponding
4104 element is non zero in live_subregs_used. The sbitmap size of
4105 live_subreg[allocno] is number of bytes that the pseudo can
4106 occupy. */
4107 sbitmap *live_subregs = XCNEWVEC (sbitmap, max_regno);
4108 bitmap live_subregs_used = BITMAP_ALLOC (NULL);
4110 for (i = 0; i < FIRST_PSEUDO_REGISTER; i++)
4111 if (TEST_HARD_REG_BIT (eliminable_regset, i))
4112 bitmap_set_bit (elim_regset, i);
4113 FOR_EACH_BB_REVERSE_FN (bb, cfun)
4115 bitmap_iterator bi;
4116 rtx_insn *insn;
4118 CLEAR_REG_SET (live_relevant_regs);
4119 bitmap_clear (live_subregs_used);
4121 EXECUTE_IF_SET_IN_BITMAP (df_get_live_out (bb), 0, i, bi)
4123 if (i >= FIRST_PSEUDO_REGISTER)
4124 break;
4125 bitmap_set_bit (live_relevant_regs, i);
4128 EXECUTE_IF_SET_IN_BITMAP (df_get_live_out (bb),
4129 FIRST_PSEUDO_REGISTER, i, bi)
4131 if (pseudo_for_reload_consideration_p (i))
4132 bitmap_set_bit (live_relevant_regs, i);
4135 FOR_BB_INSNS_REVERSE (bb, insn)
4137 if (!NOTE_P (insn) && !BARRIER_P (insn))
4139 struct df_insn_info *insn_info = DF_INSN_INFO_GET (insn);
4140 df_ref def, use;
4142 c = new_insn_chain ();
4143 c->next = next;
4144 next = c;
4145 *p = c;
4146 p = &c->prev;
4148 c->insn = insn;
4149 c->block = bb->index;
4151 if (NONDEBUG_INSN_P (insn))
4152 FOR_EACH_INSN_INFO_DEF (def, insn_info)
4154 unsigned int regno = DF_REF_REGNO (def);
4156 /* Ignore may clobbers because these are generated
4157 from calls. However, every other kind of def is
4158 added to dead_or_set. */
4159 if (!DF_REF_FLAGS_IS_SET (def, DF_REF_MAY_CLOBBER))
4161 if (regno < FIRST_PSEUDO_REGISTER)
4163 if (!fixed_regs[regno])
4164 bitmap_set_bit (&c->dead_or_set, regno);
4166 else if (pseudo_for_reload_consideration_p (regno))
4167 bitmap_set_bit (&c->dead_or_set, regno);
4170 if ((regno < FIRST_PSEUDO_REGISTER
4171 || reg_renumber[regno] >= 0
4172 || ira_conflicts_p)
4173 && (!DF_REF_FLAGS_IS_SET (def, DF_REF_CONDITIONAL)))
4175 rtx reg = DF_REF_REG (def);
4177 /* We can model subregs, but not if they are
4178 wrapped in ZERO_EXTRACTS. */
4179 if (GET_CODE (reg) == SUBREG
4180 && !DF_REF_FLAGS_IS_SET (def, DF_REF_ZERO_EXTRACT))
4182 unsigned int start = SUBREG_BYTE (reg);
4183 unsigned int last = start
4184 + GET_MODE_SIZE (GET_MODE (reg));
4186 init_live_subregs
4187 (bitmap_bit_p (live_relevant_regs, regno),
4188 live_subregs, live_subregs_used, regno, reg);
4190 if (!DF_REF_FLAGS_IS_SET
4191 (def, DF_REF_STRICT_LOW_PART))
4193 /* Expand the range to cover entire words.
4194 Bytes added here are "don't care". */
4195 start
4196 = start / UNITS_PER_WORD * UNITS_PER_WORD;
4197 last = ((last + UNITS_PER_WORD - 1)
4198 / UNITS_PER_WORD * UNITS_PER_WORD);
4201 /* Ignore the paradoxical bits. */
4202 if (last > SBITMAP_SIZE (live_subregs[regno]))
4203 last = SBITMAP_SIZE (live_subregs[regno]);
4205 while (start < last)
4207 bitmap_clear_bit (live_subregs[regno], start);
4208 start++;
4211 if (bitmap_empty_p (live_subregs[regno]))
4213 bitmap_clear_bit (live_subregs_used, regno);
4214 bitmap_clear_bit (live_relevant_regs, regno);
4216 else
4217 /* Set live_relevant_regs here because
4218 that bit has to be true to get us to
4219 look at the live_subregs fields. */
4220 bitmap_set_bit (live_relevant_regs, regno);
4222 else
4224 /* DF_REF_PARTIAL is generated for
4225 subregs, STRICT_LOW_PART, and
4226 ZERO_EXTRACT. We handle the subreg
4227 case above so here we have to keep from
4228 modeling the def as a killing def. */
4229 if (!DF_REF_FLAGS_IS_SET (def, DF_REF_PARTIAL))
4231 bitmap_clear_bit (live_subregs_used, regno);
4232 bitmap_clear_bit (live_relevant_regs, regno);
4238 bitmap_and_compl_into (live_relevant_regs, elim_regset);
4239 bitmap_copy (&c->live_throughout, live_relevant_regs);
4241 if (NONDEBUG_INSN_P (insn))
4242 FOR_EACH_INSN_INFO_USE (use, insn_info)
4244 unsigned int regno = DF_REF_REGNO (use);
4245 rtx reg = DF_REF_REG (use);
4247 /* DF_REF_READ_WRITE on a use means that this use
4248 is fabricated from a def that is a partial set
4249 to a multiword reg. Here, we only model the
4250 subreg case that is not wrapped in ZERO_EXTRACT
4251 precisely so we do not need to look at the
4252 fabricated use. */
4253 if (DF_REF_FLAGS_IS_SET (use, DF_REF_READ_WRITE)
4254 && !DF_REF_FLAGS_IS_SET (use, DF_REF_ZERO_EXTRACT)
4255 && DF_REF_FLAGS_IS_SET (use, DF_REF_SUBREG))
4256 continue;
4258 /* Add the last use of each var to dead_or_set. */
4259 if (!bitmap_bit_p (live_relevant_regs, regno))
4261 if (regno < FIRST_PSEUDO_REGISTER)
4263 if (!fixed_regs[regno])
4264 bitmap_set_bit (&c->dead_or_set, regno);
4266 else if (pseudo_for_reload_consideration_p (regno))
4267 bitmap_set_bit (&c->dead_or_set, regno);
4270 if (regno < FIRST_PSEUDO_REGISTER
4271 || pseudo_for_reload_consideration_p (regno))
4273 if (GET_CODE (reg) == SUBREG
4274 && !DF_REF_FLAGS_IS_SET (use,
4275 DF_REF_SIGN_EXTRACT
4276 | DF_REF_ZERO_EXTRACT))
4278 unsigned int start = SUBREG_BYTE (reg);
4279 unsigned int last = start
4280 + GET_MODE_SIZE (GET_MODE (reg));
4282 init_live_subregs
4283 (bitmap_bit_p (live_relevant_regs, regno),
4284 live_subregs, live_subregs_used, regno, reg);
4286 /* Ignore the paradoxical bits. */
4287 if (last > SBITMAP_SIZE (live_subregs[regno]))
4288 last = SBITMAP_SIZE (live_subregs[regno]);
4290 while (start < last)
4292 bitmap_set_bit (live_subregs[regno], start);
4293 start++;
4296 else
4297 /* Resetting the live_subregs_used is
4298 effectively saying do not use the subregs
4299 because we are reading the whole
4300 pseudo. */
4301 bitmap_clear_bit (live_subregs_used, regno);
4302 bitmap_set_bit (live_relevant_regs, regno);
4308 /* FIXME!! The following code is a disaster. Reload needs to see the
4309 labels and jump tables that are just hanging out in between
4310 the basic blocks. See pr33676. */
4311 insn = BB_HEAD (bb);
4313 /* Skip over the barriers and cruft. */
4314 while (insn && (BARRIER_P (insn) || NOTE_P (insn)
4315 || BLOCK_FOR_INSN (insn) == bb))
4316 insn = PREV_INSN (insn);
4318 /* While we add anything except barriers and notes, the focus is
4319 to get the labels and jump tables into the
4320 reload_insn_chain. */
4321 while (insn)
4323 if (!NOTE_P (insn) && !BARRIER_P (insn))
4325 if (BLOCK_FOR_INSN (insn))
4326 break;
4328 c = new_insn_chain ();
4329 c->next = next;
4330 next = c;
4331 *p = c;
4332 p = &c->prev;
4334 /* The block makes no sense here, but it is what the old
4335 code did. */
4336 c->block = bb->index;
4337 c->insn = insn;
4338 bitmap_copy (&c->live_throughout, live_relevant_regs);
4340 insn = PREV_INSN (insn);
4344 reload_insn_chain = c;
4345 *p = NULL;
4347 for (i = 0; i < (unsigned int) max_regno; i++)
4348 if (live_subregs[i] != NULL)
4349 sbitmap_free (live_subregs[i]);
4350 free (live_subregs);
4351 BITMAP_FREE (live_subregs_used);
4352 BITMAP_FREE (live_relevant_regs);
4353 BITMAP_FREE (elim_regset);
4355 if (dump_file)
4356 print_insn_chains (dump_file);
4359 /* Examine the rtx found in *LOC, which is read or written to as determined
4360 by TYPE. Return false if we find a reason why an insn containing this
4361 rtx should not be moved (such as accesses to non-constant memory), true
4362 otherwise. */
4363 static bool
4364 rtx_moveable_p (rtx *loc, enum op_type type)
4366 const char *fmt;
4367 rtx x = *loc;
4368 enum rtx_code code = GET_CODE (x);
4369 int i, j;
4371 code = GET_CODE (x);
4372 switch (code)
4374 case CONST:
4375 CASE_CONST_ANY:
4376 case SYMBOL_REF:
4377 case LABEL_REF:
4378 return true;
4380 case PC:
4381 return type == OP_IN;
4383 case CC0:
4384 return false;
4386 case REG:
4387 if (x == frame_pointer_rtx)
4388 return true;
4389 if (HARD_REGISTER_P (x))
4390 return false;
4392 return true;
4394 case MEM:
4395 if (type == OP_IN && MEM_READONLY_P (x))
4396 return rtx_moveable_p (&XEXP (x, 0), OP_IN);
4397 return false;
4399 case SET:
4400 return (rtx_moveable_p (&SET_SRC (x), OP_IN)
4401 && rtx_moveable_p (&SET_DEST (x), OP_OUT));
4403 case STRICT_LOW_PART:
4404 return rtx_moveable_p (&XEXP (x, 0), OP_OUT);
4406 case ZERO_EXTRACT:
4407 case SIGN_EXTRACT:
4408 return (rtx_moveable_p (&XEXP (x, 0), type)
4409 && rtx_moveable_p (&XEXP (x, 1), OP_IN)
4410 && rtx_moveable_p (&XEXP (x, 2), OP_IN));
4412 case CLOBBER:
4413 return rtx_moveable_p (&SET_DEST (x), OP_OUT);
4415 case UNSPEC_VOLATILE:
4416 /* It is a bad idea to consider insns with with such rtl
4417 as moveable ones. The insn scheduler also considers them as barrier
4418 for a reason. */
4419 return false;
4421 default:
4422 break;
4425 fmt = GET_RTX_FORMAT (code);
4426 for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
4428 if (fmt[i] == 'e')
4430 if (!rtx_moveable_p (&XEXP (x, i), type))
4431 return false;
4433 else if (fmt[i] == 'E')
4434 for (j = XVECLEN (x, i) - 1; j >= 0; j--)
4436 if (!rtx_moveable_p (&XVECEXP (x, i, j), type))
4437 return false;
4440 return true;
4443 /* A wrapper around dominated_by_p, which uses the information in UID_LUID
4444 to give dominance relationships between two insns I1 and I2. */
4445 static bool
4446 insn_dominated_by_p (rtx i1, rtx i2, int *uid_luid)
4448 basic_block bb1 = BLOCK_FOR_INSN (i1);
4449 basic_block bb2 = BLOCK_FOR_INSN (i2);
4451 if (bb1 == bb2)
4452 return uid_luid[INSN_UID (i2)] < uid_luid[INSN_UID (i1)];
4453 return dominated_by_p (CDI_DOMINATORS, bb1, bb2);
4456 /* Record the range of register numbers added by find_moveable_pseudos. */
4457 int first_moveable_pseudo, last_moveable_pseudo;
4459 /* These two vectors hold data for every register added by
4460 find_movable_pseudos, with index 0 holding data for the
4461 first_moveable_pseudo. */
4462 /* The original home register. */
4463 static vec<rtx> pseudo_replaced_reg;
4465 /* Look for instances where we have an instruction that is known to increase
4466 register pressure, and whose result is not used immediately. If it is
4467 possible to move the instruction downwards to just before its first use,
4468 split its lifetime into two ranges. We create a new pseudo to compute the
4469 value, and emit a move instruction just before the first use. If, after
4470 register allocation, the new pseudo remains unallocated, the function
4471 move_unallocated_pseudos then deletes the move instruction and places
4472 the computation just before the first use.
4474 Such a move is safe and profitable if all the input registers remain live
4475 and unchanged between the original computation and its first use. In such
4476 a situation, the computation is known to increase register pressure, and
4477 moving it is known to at least not worsen it.
4479 We restrict moves to only those cases where a register remains unallocated,
4480 in order to avoid interfering too much with the instruction schedule. As
4481 an exception, we may move insns which only modify their input register
4482 (typically induction variables), as this increases the freedom for our
4483 intended transformation, and does not limit the second instruction
4484 scheduler pass. */
4486 static void
4487 find_moveable_pseudos (void)
4489 unsigned i;
4490 int max_regs = max_reg_num ();
4491 int max_uid = get_max_uid ();
4492 basic_block bb;
4493 int *uid_luid = XNEWVEC (int, max_uid);
4494 rtx_insn **closest_uses = XNEWVEC (rtx_insn *, max_regs);
4495 /* A set of registers which are live but not modified throughout a block. */
4496 bitmap_head *bb_transp_live = XNEWVEC (bitmap_head,
4497 last_basic_block_for_fn (cfun));
4498 /* A set of registers which only exist in a given basic block. */
4499 bitmap_head *bb_local = XNEWVEC (bitmap_head,
4500 last_basic_block_for_fn (cfun));
4501 /* A set of registers which are set once, in an instruction that can be
4502 moved freely downwards, but are otherwise transparent to a block. */
4503 bitmap_head *bb_moveable_reg_sets = XNEWVEC (bitmap_head,
4504 last_basic_block_for_fn (cfun));
4505 bitmap_head live, used, set, interesting, unusable_as_input;
4506 bitmap_iterator bi;
4507 bitmap_initialize (&interesting, 0);
4509 first_moveable_pseudo = max_regs;
4510 pseudo_replaced_reg.release ();
4511 pseudo_replaced_reg.safe_grow_cleared (max_regs);
4513 df_analyze ();
4514 calculate_dominance_info (CDI_DOMINATORS);
4516 i = 0;
4517 bitmap_initialize (&live, 0);
4518 bitmap_initialize (&used, 0);
4519 bitmap_initialize (&set, 0);
4520 bitmap_initialize (&unusable_as_input, 0);
4521 FOR_EACH_BB_FN (bb, cfun)
4523 rtx_insn *insn;
4524 bitmap transp = bb_transp_live + bb->index;
4525 bitmap moveable = bb_moveable_reg_sets + bb->index;
4526 bitmap local = bb_local + bb->index;
4528 bitmap_initialize (local, 0);
4529 bitmap_initialize (transp, 0);
4530 bitmap_initialize (moveable, 0);
4531 bitmap_copy (&live, df_get_live_out (bb));
4532 bitmap_and_into (&live, df_get_live_in (bb));
4533 bitmap_copy (transp, &live);
4534 bitmap_clear (moveable);
4535 bitmap_clear (&live);
4536 bitmap_clear (&used);
4537 bitmap_clear (&set);
4538 FOR_BB_INSNS (bb, insn)
4539 if (NONDEBUG_INSN_P (insn))
4541 df_insn_info *insn_info = DF_INSN_INFO_GET (insn);
4542 df_ref def, use;
4544 uid_luid[INSN_UID (insn)] = i++;
4546 def = df_single_def (insn_info);
4547 use = df_single_use (insn_info);
4548 if (use
4549 && def
4550 && DF_REF_REGNO (use) == DF_REF_REGNO (def)
4551 && !bitmap_bit_p (&set, DF_REF_REGNO (use))
4552 && rtx_moveable_p (&PATTERN (insn), OP_IN))
4554 unsigned regno = DF_REF_REGNO (use);
4555 bitmap_set_bit (moveable, regno);
4556 bitmap_set_bit (&set, regno);
4557 bitmap_set_bit (&used, regno);
4558 bitmap_clear_bit (transp, regno);
4559 continue;
4561 FOR_EACH_INSN_INFO_USE (use, insn_info)
4563 unsigned regno = DF_REF_REGNO (use);
4564 bitmap_set_bit (&used, regno);
4565 if (bitmap_clear_bit (moveable, regno))
4566 bitmap_clear_bit (transp, regno);
4569 FOR_EACH_INSN_INFO_DEF (def, insn_info)
4571 unsigned regno = DF_REF_REGNO (def);
4572 bitmap_set_bit (&set, regno);
4573 bitmap_clear_bit (transp, regno);
4574 bitmap_clear_bit (moveable, regno);
4579 bitmap_clear (&live);
4580 bitmap_clear (&used);
4581 bitmap_clear (&set);
4583 FOR_EACH_BB_FN (bb, cfun)
4585 bitmap local = bb_local + bb->index;
4586 rtx_insn *insn;
4588 FOR_BB_INSNS (bb, insn)
4589 if (NONDEBUG_INSN_P (insn))
4591 df_insn_info *insn_info = DF_INSN_INFO_GET (insn);
4592 rtx_insn *def_insn;
4593 rtx closest_use, note;
4594 df_ref def, use;
4595 unsigned regno;
4596 bool all_dominated, all_local;
4597 machine_mode mode;
4599 def = df_single_def (insn_info);
4600 /* There must be exactly one def in this insn. */
4601 if (!def || !single_set (insn))
4602 continue;
4603 /* This must be the only definition of the reg. We also limit
4604 which modes we deal with so that we can assume we can generate
4605 move instructions. */
4606 regno = DF_REF_REGNO (def);
4607 mode = GET_MODE (DF_REF_REG (def));
4608 if (DF_REG_DEF_COUNT (regno) != 1
4609 || !DF_REF_INSN_INFO (def)
4610 || HARD_REGISTER_NUM_P (regno)
4611 || DF_REG_EQ_USE_COUNT (regno) > 0
4612 || (!INTEGRAL_MODE_P (mode) && !FLOAT_MODE_P (mode)))
4613 continue;
4614 def_insn = DF_REF_INSN (def);
4616 for (note = REG_NOTES (def_insn); note; note = XEXP (note, 1))
4617 if (REG_NOTE_KIND (note) == REG_EQUIV && MEM_P (XEXP (note, 0)))
4618 break;
4620 if (note)
4622 if (dump_file)
4623 fprintf (dump_file, "Ignoring reg %d, has equiv memory\n",
4624 regno);
4625 bitmap_set_bit (&unusable_as_input, regno);
4626 continue;
4629 use = DF_REG_USE_CHAIN (regno);
4630 all_dominated = true;
4631 all_local = true;
4632 closest_use = NULL_RTX;
4633 for (; use; use = DF_REF_NEXT_REG (use))
4635 rtx_insn *insn;
4636 if (!DF_REF_INSN_INFO (use))
4638 all_dominated = false;
4639 all_local = false;
4640 break;
4642 insn = DF_REF_INSN (use);
4643 if (DEBUG_INSN_P (insn))
4644 continue;
4645 if (BLOCK_FOR_INSN (insn) != BLOCK_FOR_INSN (def_insn))
4646 all_local = false;
4647 if (!insn_dominated_by_p (insn, def_insn, uid_luid))
4648 all_dominated = false;
4649 if (closest_use != insn && closest_use != const0_rtx)
4651 if (closest_use == NULL_RTX)
4652 closest_use = insn;
4653 else if (insn_dominated_by_p (closest_use, insn, uid_luid))
4654 closest_use = insn;
4655 else if (!insn_dominated_by_p (insn, closest_use, uid_luid))
4656 closest_use = const0_rtx;
4659 if (!all_dominated)
4661 if (dump_file)
4662 fprintf (dump_file, "Reg %d not all uses dominated by set\n",
4663 regno);
4664 continue;
4666 if (all_local)
4667 bitmap_set_bit (local, regno);
4668 if (closest_use == const0_rtx || closest_use == NULL
4669 || next_nonnote_nondebug_insn (def_insn) == closest_use)
4671 if (dump_file)
4672 fprintf (dump_file, "Reg %d uninteresting%s\n", regno,
4673 closest_use == const0_rtx || closest_use == NULL
4674 ? " (no unique first use)" : "");
4675 continue;
4677 #ifdef HAVE_cc0
4678 if (reg_referenced_p (cc0_rtx, PATTERN (closest_use)))
4680 if (dump_file)
4681 fprintf (dump_file, "Reg %d: closest user uses cc0\n",
4682 regno);
4683 continue;
4685 #endif
4686 bitmap_set_bit (&interesting, regno);
4687 /* If we get here, we know closest_use is a non-NULL insn
4688 (as opposed to const_0_rtx). */
4689 closest_uses[regno] = as_a <rtx_insn *> (closest_use);
4691 if (dump_file && (all_local || all_dominated))
4693 fprintf (dump_file, "Reg %u:", regno);
4694 if (all_local)
4695 fprintf (dump_file, " local to bb %d", bb->index);
4696 if (all_dominated)
4697 fprintf (dump_file, " def dominates all uses");
4698 if (closest_use != const0_rtx)
4699 fprintf (dump_file, " has unique first use");
4700 fputs ("\n", dump_file);
4705 EXECUTE_IF_SET_IN_BITMAP (&interesting, 0, i, bi)
4707 df_ref def = DF_REG_DEF_CHAIN (i);
4708 rtx_insn *def_insn = DF_REF_INSN (def);
4709 basic_block def_block = BLOCK_FOR_INSN (def_insn);
4710 bitmap def_bb_local = bb_local + def_block->index;
4711 bitmap def_bb_moveable = bb_moveable_reg_sets + def_block->index;
4712 bitmap def_bb_transp = bb_transp_live + def_block->index;
4713 bool local_to_bb_p = bitmap_bit_p (def_bb_local, i);
4714 rtx_insn *use_insn = closest_uses[i];
4715 df_ref use;
4716 bool all_ok = true;
4717 bool all_transp = true;
4719 if (!REG_P (DF_REF_REG (def)))
4720 continue;
4722 if (!local_to_bb_p)
4724 if (dump_file)
4725 fprintf (dump_file, "Reg %u not local to one basic block\n",
4727 continue;
4729 if (reg_equiv_init (i) != NULL_RTX)
4731 if (dump_file)
4732 fprintf (dump_file, "Ignoring reg %u with equiv init insn\n",
4734 continue;
4736 if (!rtx_moveable_p (&PATTERN (def_insn), OP_IN))
4738 if (dump_file)
4739 fprintf (dump_file, "Found def insn %d for %d to be not moveable\n",
4740 INSN_UID (def_insn), i);
4741 continue;
4743 if (dump_file)
4744 fprintf (dump_file, "Examining insn %d, def for %d\n",
4745 INSN_UID (def_insn), i);
4746 FOR_EACH_INSN_USE (use, def_insn)
4748 unsigned regno = DF_REF_REGNO (use);
4749 if (bitmap_bit_p (&unusable_as_input, regno))
4751 all_ok = false;
4752 if (dump_file)
4753 fprintf (dump_file, " found unusable input reg %u.\n", regno);
4754 break;
4756 if (!bitmap_bit_p (def_bb_transp, regno))
4758 if (bitmap_bit_p (def_bb_moveable, regno)
4759 && !control_flow_insn_p (use_insn)
4760 #ifdef HAVE_cc0
4761 && !sets_cc0_p (use_insn)
4762 #endif
4765 if (modified_between_p (DF_REF_REG (use), def_insn, use_insn))
4767 rtx_insn *x = NEXT_INSN (def_insn);
4768 while (!modified_in_p (DF_REF_REG (use), x))
4770 gcc_assert (x != use_insn);
4771 x = NEXT_INSN (x);
4773 if (dump_file)
4774 fprintf (dump_file, " input reg %u modified but insn %d moveable\n",
4775 regno, INSN_UID (x));
4776 emit_insn_after (PATTERN (x), use_insn);
4777 set_insn_deleted (x);
4779 else
4781 if (dump_file)
4782 fprintf (dump_file, " input reg %u modified between def and use\n",
4783 regno);
4784 all_transp = false;
4787 else
4788 all_transp = false;
4791 if (!all_ok)
4792 continue;
4793 if (!dbg_cnt (ira_move))
4794 break;
4795 if (dump_file)
4796 fprintf (dump_file, " all ok%s\n", all_transp ? " and transp" : "");
4798 if (all_transp)
4800 rtx def_reg = DF_REF_REG (def);
4801 rtx newreg = ira_create_new_reg (def_reg);
4802 if (validate_change (def_insn, DF_REF_REAL_LOC (def), newreg, 0))
4804 unsigned nregno = REGNO (newreg);
4805 emit_insn_before (gen_move_insn (def_reg, newreg), use_insn);
4806 nregno -= max_regs;
4807 pseudo_replaced_reg[nregno] = def_reg;
4812 FOR_EACH_BB_FN (bb, cfun)
4814 bitmap_clear (bb_local + bb->index);
4815 bitmap_clear (bb_transp_live + bb->index);
4816 bitmap_clear (bb_moveable_reg_sets + bb->index);
4818 bitmap_clear (&interesting);
4819 bitmap_clear (&unusable_as_input);
4820 free (uid_luid);
4821 free (closest_uses);
4822 free (bb_local);
4823 free (bb_transp_live);
4824 free (bb_moveable_reg_sets);
4826 last_moveable_pseudo = max_reg_num ();
4828 fix_reg_equiv_init ();
4829 expand_reg_info ();
4830 regstat_free_n_sets_and_refs ();
4831 regstat_free_ri ();
4832 regstat_init_n_sets_and_refs ();
4833 regstat_compute_ri ();
4834 free_dominance_info (CDI_DOMINATORS);
4837 /* If SET pattern SET is an assignment from a hard register to a pseudo which
4838 is live at CALL_DOM (if non-NULL, otherwise this check is omitted), return
4839 the destination. Otherwise return NULL. */
4841 static rtx
4842 interesting_dest_for_shprep_1 (rtx set, basic_block call_dom)
4844 rtx src = SET_SRC (set);
4845 rtx dest = SET_DEST (set);
4846 if (!REG_P (src) || !HARD_REGISTER_P (src)
4847 || !REG_P (dest) || HARD_REGISTER_P (dest)
4848 || (call_dom && !bitmap_bit_p (df_get_live_in (call_dom), REGNO (dest))))
4849 return NULL;
4850 return dest;
4853 /* If insn is interesting for parameter range-splitting shrink-wrapping
4854 preparation, i.e. it is a single set from a hard register to a pseudo, which
4855 is live at CALL_DOM (if non-NULL, otherwise this check is omitted), or a
4856 parallel statement with only one such statement, return the destination.
4857 Otherwise return NULL. */
4859 static rtx
4860 interesting_dest_for_shprep (rtx_insn *insn, basic_block call_dom)
4862 if (!INSN_P (insn))
4863 return NULL;
4864 rtx pat = PATTERN (insn);
4865 if (GET_CODE (pat) == SET)
4866 return interesting_dest_for_shprep_1 (pat, call_dom);
4868 if (GET_CODE (pat) != PARALLEL)
4869 return NULL;
4870 rtx ret = NULL;
4871 for (int i = 0; i < XVECLEN (pat, 0); i++)
4873 rtx sub = XVECEXP (pat, 0, i);
4874 if (GET_CODE (sub) == USE || GET_CODE (sub) == CLOBBER)
4875 continue;
4876 if (GET_CODE (sub) != SET
4877 || side_effects_p (sub))
4878 return NULL;
4879 rtx dest = interesting_dest_for_shprep_1 (sub, call_dom);
4880 if (dest && ret)
4881 return NULL;
4882 if (dest)
4883 ret = dest;
4885 return ret;
4888 /* Split live ranges of pseudos that are loaded from hard registers in the
4889 first BB in a BB that dominates all non-sibling call if such a BB can be
4890 found and is not in a loop. Return true if the function has made any
4891 changes. */
4893 static bool
4894 split_live_ranges_for_shrink_wrap (void)
4896 basic_block bb, call_dom = NULL;
4897 basic_block first = single_succ (ENTRY_BLOCK_PTR_FOR_FN (cfun));
4898 rtx_insn *insn, *last_interesting_insn = NULL;
4899 bitmap_head need_new, reachable;
4900 vec<basic_block> queue;
4902 if (!SHRINK_WRAPPING_ENABLED)
4903 return false;
4905 bitmap_initialize (&need_new, 0);
4906 bitmap_initialize (&reachable, 0);
4907 queue.create (n_basic_blocks_for_fn (cfun));
4909 FOR_EACH_BB_FN (bb, cfun)
4910 FOR_BB_INSNS (bb, insn)
4911 if (CALL_P (insn) && !SIBLING_CALL_P (insn))
4913 if (bb == first)
4915 bitmap_clear (&need_new);
4916 bitmap_clear (&reachable);
4917 queue.release ();
4918 return false;
4921 bitmap_set_bit (&need_new, bb->index);
4922 bitmap_set_bit (&reachable, bb->index);
4923 queue.quick_push (bb);
4924 break;
4927 if (queue.is_empty ())
4929 bitmap_clear (&need_new);
4930 bitmap_clear (&reachable);
4931 queue.release ();
4932 return false;
4935 while (!queue.is_empty ())
4937 edge e;
4938 edge_iterator ei;
4940 bb = queue.pop ();
4941 FOR_EACH_EDGE (e, ei, bb->succs)
4942 if (e->dest != EXIT_BLOCK_PTR_FOR_FN (cfun)
4943 && bitmap_set_bit (&reachable, e->dest->index))
4944 queue.quick_push (e->dest);
4946 queue.release ();
4948 FOR_BB_INSNS (first, insn)
4950 rtx dest = interesting_dest_for_shprep (insn, NULL);
4951 if (!dest)
4952 continue;
4954 if (DF_REG_DEF_COUNT (REGNO (dest)) > 1)
4956 bitmap_clear (&need_new);
4957 bitmap_clear (&reachable);
4958 return false;
4961 for (df_ref use = DF_REG_USE_CHAIN (REGNO(dest));
4962 use;
4963 use = DF_REF_NEXT_REG (use))
4965 int ubbi = DF_REF_BB (use)->index;
4966 if (bitmap_bit_p (&reachable, ubbi))
4967 bitmap_set_bit (&need_new, ubbi);
4969 last_interesting_insn = insn;
4972 bitmap_clear (&reachable);
4973 if (!last_interesting_insn)
4975 bitmap_clear (&need_new);
4976 return false;
4979 call_dom = nearest_common_dominator_for_set (CDI_DOMINATORS, &need_new);
4980 bitmap_clear (&need_new);
4981 if (call_dom == first)
4982 return false;
4984 loop_optimizer_init (AVOID_CFG_MODIFICATIONS);
4985 while (bb_loop_depth (call_dom) > 0)
4986 call_dom = get_immediate_dominator (CDI_DOMINATORS, call_dom);
4987 loop_optimizer_finalize ();
4989 if (call_dom == first)
4990 return false;
4992 calculate_dominance_info (CDI_POST_DOMINATORS);
4993 if (dominated_by_p (CDI_POST_DOMINATORS, first, call_dom))
4995 free_dominance_info (CDI_POST_DOMINATORS);
4996 return false;
4998 free_dominance_info (CDI_POST_DOMINATORS);
5000 if (dump_file)
5001 fprintf (dump_file, "Will split live ranges of parameters at BB %i\n",
5002 call_dom->index);
5004 bool ret = false;
5005 FOR_BB_INSNS (first, insn)
5007 rtx dest = interesting_dest_for_shprep (insn, call_dom);
5008 if (!dest || dest == pic_offset_table_rtx)
5009 continue;
5011 rtx newreg = NULL_RTX;
5012 df_ref use, next;
5013 for (use = DF_REG_USE_CHAIN (REGNO (dest)); use; use = next)
5015 rtx_insn *uin = DF_REF_INSN (use);
5016 next = DF_REF_NEXT_REG (use);
5018 basic_block ubb = BLOCK_FOR_INSN (uin);
5019 if (ubb == call_dom
5020 || dominated_by_p (CDI_DOMINATORS, ubb, call_dom))
5022 if (!newreg)
5023 newreg = ira_create_new_reg (dest);
5024 validate_change (uin, DF_REF_REAL_LOC (use), newreg, true);
5028 if (newreg)
5030 rtx new_move = gen_move_insn (newreg, dest);
5031 emit_insn_after (new_move, bb_note (call_dom));
5032 if (dump_file)
5034 fprintf (dump_file, "Split live-range of register ");
5035 print_rtl_single (dump_file, dest);
5037 ret = true;
5040 if (insn == last_interesting_insn)
5041 break;
5043 apply_change_group ();
5044 return ret;
5047 /* Perform the second half of the transformation started in
5048 find_moveable_pseudos. We look for instances where the newly introduced
5049 pseudo remains unallocated, and remove it by moving the definition to
5050 just before its use, replacing the move instruction generated by
5051 find_moveable_pseudos. */
5052 static void
5053 move_unallocated_pseudos (void)
5055 int i;
5056 for (i = first_moveable_pseudo; i < last_moveable_pseudo; i++)
5057 if (reg_renumber[i] < 0)
5059 int idx = i - first_moveable_pseudo;
5060 rtx other_reg = pseudo_replaced_reg[idx];
5061 rtx_insn *def_insn = DF_REF_INSN (DF_REG_DEF_CHAIN (i));
5062 /* The use must follow all definitions of OTHER_REG, so we can
5063 insert the new definition immediately after any of them. */
5064 df_ref other_def = DF_REG_DEF_CHAIN (REGNO (other_reg));
5065 rtx_insn *move_insn = DF_REF_INSN (other_def);
5066 rtx_insn *newinsn = emit_insn_after (PATTERN (def_insn), move_insn);
5067 rtx set;
5068 int success;
5070 if (dump_file)
5071 fprintf (dump_file, "moving def of %d (insn %d now) ",
5072 REGNO (other_reg), INSN_UID (def_insn));
5074 delete_insn (move_insn);
5075 while ((other_def = DF_REG_DEF_CHAIN (REGNO (other_reg))))
5076 delete_insn (DF_REF_INSN (other_def));
5077 delete_insn (def_insn);
5079 set = single_set (newinsn);
5080 success = validate_change (newinsn, &SET_DEST (set), other_reg, 0);
5081 gcc_assert (success);
5082 if (dump_file)
5083 fprintf (dump_file, " %d) rather than keep unallocated replacement %d\n",
5084 INSN_UID (newinsn), i);
5085 SET_REG_N_REFS (i, 0);
5089 /* If the backend knows where to allocate pseudos for hard
5090 register initial values, register these allocations now. */
5091 static void
5092 allocate_initial_values (void)
5094 if (targetm.allocate_initial_value)
5096 rtx hreg, preg, x;
5097 int i, regno;
5099 for (i = 0; HARD_REGISTER_NUM_P (i); i++)
5101 if (! initial_value_entry (i, &hreg, &preg))
5102 break;
5104 x = targetm.allocate_initial_value (hreg);
5105 regno = REGNO (preg);
5106 if (x && REG_N_SETS (regno) <= 1)
5108 if (MEM_P (x))
5109 reg_equiv_memory_loc (regno) = x;
5110 else
5112 basic_block bb;
5113 int new_regno;
5115 gcc_assert (REG_P (x));
5116 new_regno = REGNO (x);
5117 reg_renumber[regno] = new_regno;
5118 /* Poke the regno right into regno_reg_rtx so that even
5119 fixed regs are accepted. */
5120 SET_REGNO (preg, new_regno);
5121 /* Update global register liveness information. */
5122 FOR_EACH_BB_FN (bb, cfun)
5124 if (REGNO_REG_SET_P (df_get_live_in (bb), regno))
5125 SET_REGNO_REG_SET (df_get_live_in (bb), new_regno);
5126 if (REGNO_REG_SET_P (df_get_live_out (bb), regno))
5127 SET_REGNO_REG_SET (df_get_live_out (bb), new_regno);
5133 gcc_checking_assert (! initial_value_entry (FIRST_PSEUDO_REGISTER,
5134 &hreg, &preg));
5139 /* True when we use LRA instead of reload pass for the current
5140 function. */
5141 bool ira_use_lra_p;
5143 /* True if we have allocno conflicts. It is false for non-optimized
5144 mode or when the conflict table is too big. */
5145 bool ira_conflicts_p;
5147 /* Saved between IRA and reload. */
5148 static int saved_flag_ira_share_spill_slots;
5150 /* This is the main entry of IRA. */
5151 static void
5152 ira (FILE *f)
5154 bool loops_p;
5155 int ira_max_point_before_emit;
5156 bool saved_flag_caller_saves = flag_caller_saves;
5157 enum ira_region saved_flag_ira_region = flag_ira_region;
5159 /* Perform target specific PIC register initialization. */
5160 targetm.init_pic_reg ();
5162 ira_conflicts_p = optimize > 0;
5164 ira_use_lra_p = targetm.lra_p ();
5165 /* If there are too many pseudos and/or basic blocks (e.g. 10K
5166 pseudos and 10K blocks or 100K pseudos and 1K blocks), we will
5167 use simplified and faster algorithms in LRA. */
5168 lra_simple_p
5169 = (ira_use_lra_p
5170 && max_reg_num () >= (1 << 26) / last_basic_block_for_fn (cfun));
5171 if (lra_simple_p)
5173 /* It permits to skip live range splitting in LRA. */
5174 flag_caller_saves = false;
5175 /* There is no sense to do regional allocation when we use
5176 simplified LRA. */
5177 flag_ira_region = IRA_REGION_ONE;
5178 ira_conflicts_p = false;
5181 #ifndef IRA_NO_OBSTACK
5182 gcc_obstack_init (&ira_obstack);
5183 #endif
5184 bitmap_obstack_initialize (&ira_bitmap_obstack);
5186 /* LRA uses its own infrastructure to handle caller save registers. */
5187 if (flag_caller_saves && !ira_use_lra_p)
5188 init_caller_save ();
5190 if (flag_ira_verbose < 10)
5192 internal_flag_ira_verbose = flag_ira_verbose;
5193 ira_dump_file = f;
5195 else
5197 internal_flag_ira_verbose = flag_ira_verbose - 10;
5198 ira_dump_file = stderr;
5201 setup_prohibited_mode_move_regs ();
5202 decrease_live_ranges_number ();
5203 df_note_add_problem ();
5205 /* DF_LIVE can't be used in the register allocator, too many other
5206 parts of the compiler depend on using the "classic" liveness
5207 interpretation of the DF_LR problem. See PR38711.
5208 Remove the problem, so that we don't spend time updating it in
5209 any of the df_analyze() calls during IRA/LRA. */
5210 if (optimize > 1)
5211 df_remove_problem (df_live);
5212 gcc_checking_assert (df_live == NULL);
5214 #ifdef ENABLE_CHECKING
5215 df->changeable_flags |= DF_VERIFY_SCHEDULED;
5216 #endif
5217 df_analyze ();
5219 init_reg_equiv ();
5220 if (ira_conflicts_p)
5222 calculate_dominance_info (CDI_DOMINATORS);
5224 if (split_live_ranges_for_shrink_wrap ())
5225 df_analyze ();
5227 free_dominance_info (CDI_DOMINATORS);
5230 df_clear_flags (DF_NO_INSN_RESCAN);
5232 indirect_jump_optimize ();
5233 if (delete_trivially_dead_insns (get_insns (), max_reg_num ()))
5234 df_analyze ();
5236 regstat_init_n_sets_and_refs ();
5237 regstat_compute_ri ();
5239 /* If we are not optimizing, then this is the only place before
5240 register allocation where dataflow is done. And that is needed
5241 to generate these warnings. */
5242 if (warn_clobbered)
5243 generate_setjmp_warnings ();
5245 /* Determine if the current function is a leaf before running IRA
5246 since this can impact optimizations done by the prologue and
5247 epilogue thus changing register elimination offsets. */
5248 crtl->is_leaf = leaf_function_p ();
5250 if (resize_reg_info () && flag_ira_loop_pressure)
5251 ira_set_pseudo_classes (true, ira_dump_file);
5253 update_equiv_regs ();
5254 setup_reg_equiv ();
5255 setup_reg_equiv_init ();
5257 allocated_reg_info_size = max_reg_num ();
5259 /* It is not worth to do such improvement when we use a simple
5260 allocation because of -O0 usage or because the function is too
5261 big. */
5262 if (ira_conflicts_p)
5263 find_moveable_pseudos ();
5265 max_regno_before_ira = max_reg_num ();
5266 ira_setup_eliminable_regset ();
5268 ira_overall_cost = ira_reg_cost = ira_mem_cost = 0;
5269 ira_load_cost = ira_store_cost = ira_shuffle_cost = 0;
5270 ira_move_loops_num = ira_additional_jumps_num = 0;
5272 ira_assert (current_loops == NULL);
5273 if (flag_ira_region == IRA_REGION_ALL || flag_ira_region == IRA_REGION_MIXED)
5274 loop_optimizer_init (AVOID_CFG_MODIFICATIONS | LOOPS_HAVE_RECORDED_EXITS);
5276 if (internal_flag_ira_verbose > 0 && ira_dump_file != NULL)
5277 fprintf (ira_dump_file, "Building IRA IR\n");
5278 loops_p = ira_build ();
5280 ira_assert (ira_conflicts_p || !loops_p);
5282 saved_flag_ira_share_spill_slots = flag_ira_share_spill_slots;
5283 if (too_high_register_pressure_p () || cfun->calls_setjmp)
5284 /* It is just wasting compiler's time to pack spilled pseudos into
5285 stack slots in this case -- prohibit it. We also do this if
5286 there is setjmp call because a variable not modified between
5287 setjmp and longjmp the compiler is required to preserve its
5288 value and sharing slots does not guarantee it. */
5289 flag_ira_share_spill_slots = FALSE;
5291 ira_color ();
5293 ira_max_point_before_emit = ira_max_point;
5295 ira_initiate_emit_data ();
5297 ira_emit (loops_p);
5299 max_regno = max_reg_num ();
5300 if (ira_conflicts_p)
5302 if (! loops_p)
5304 if (! ira_use_lra_p)
5305 ira_initiate_assign ();
5307 else
5309 expand_reg_info ();
5311 if (ira_use_lra_p)
5313 ira_allocno_t a;
5314 ira_allocno_iterator ai;
5316 FOR_EACH_ALLOCNO (a, ai)
5318 int old_regno = ALLOCNO_REGNO (a);
5319 int new_regno = REGNO (ALLOCNO_EMIT_DATA (a)->reg);
5321 ALLOCNO_REGNO (a) = new_regno;
5323 if (old_regno != new_regno)
5324 setup_reg_classes (new_regno, reg_preferred_class (old_regno),
5325 reg_alternate_class (old_regno),
5326 reg_allocno_class (old_regno));
5330 else
5332 if (internal_flag_ira_verbose > 0 && ira_dump_file != NULL)
5333 fprintf (ira_dump_file, "Flattening IR\n");
5334 ira_flattening (max_regno_before_ira, ira_max_point_before_emit);
5336 /* New insns were generated: add notes and recalculate live
5337 info. */
5338 df_analyze ();
5340 /* ??? Rebuild the loop tree, but why? Does the loop tree
5341 change if new insns were generated? Can that be handled
5342 by updating the loop tree incrementally? */
5343 loop_optimizer_finalize ();
5344 free_dominance_info (CDI_DOMINATORS);
5345 loop_optimizer_init (AVOID_CFG_MODIFICATIONS
5346 | LOOPS_HAVE_RECORDED_EXITS);
5348 if (! ira_use_lra_p)
5350 setup_allocno_assignment_flags ();
5351 ira_initiate_assign ();
5352 ira_reassign_conflict_allocnos (max_regno);
5357 ira_finish_emit_data ();
5359 setup_reg_renumber ();
5361 calculate_allocation_cost ();
5363 #ifdef ENABLE_IRA_CHECKING
5364 if (ira_conflicts_p)
5365 check_allocation ();
5366 #endif
5368 if (max_regno != max_regno_before_ira)
5370 regstat_free_n_sets_and_refs ();
5371 regstat_free_ri ();
5372 regstat_init_n_sets_and_refs ();
5373 regstat_compute_ri ();
5376 overall_cost_before = ira_overall_cost;
5377 if (! ira_conflicts_p)
5378 grow_reg_equivs ();
5379 else
5381 fix_reg_equiv_init ();
5383 #ifdef ENABLE_IRA_CHECKING
5384 print_redundant_copies ();
5385 #endif
5386 if (! ira_use_lra_p)
5388 ira_spilled_reg_stack_slots_num = 0;
5389 ira_spilled_reg_stack_slots
5390 = ((struct ira_spilled_reg_stack_slot *)
5391 ira_allocate (max_regno
5392 * sizeof (struct ira_spilled_reg_stack_slot)));
5393 memset (ira_spilled_reg_stack_slots, 0,
5394 max_regno * sizeof (struct ira_spilled_reg_stack_slot));
5397 allocate_initial_values ();
5399 /* See comment for find_moveable_pseudos call. */
5400 if (ira_conflicts_p)
5401 move_unallocated_pseudos ();
5403 /* Restore original values. */
5404 if (lra_simple_p)
5406 flag_caller_saves = saved_flag_caller_saves;
5407 flag_ira_region = saved_flag_ira_region;
5411 static void
5412 do_reload (void)
5414 basic_block bb;
5415 bool need_dce;
5416 unsigned pic_offset_table_regno = INVALID_REGNUM;
5418 if (flag_ira_verbose < 10)
5419 ira_dump_file = dump_file;
5421 /* If pic_offset_table_rtx is a pseudo register, then keep it so
5422 after reload to avoid possible wrong usages of hard reg assigned
5423 to it. */
5424 if (pic_offset_table_rtx
5425 && REGNO (pic_offset_table_rtx) >= FIRST_PSEUDO_REGISTER)
5426 pic_offset_table_regno = REGNO (pic_offset_table_rtx);
5428 timevar_push (TV_RELOAD);
5429 if (ira_use_lra_p)
5431 if (current_loops != NULL)
5433 loop_optimizer_finalize ();
5434 free_dominance_info (CDI_DOMINATORS);
5436 FOR_ALL_BB_FN (bb, cfun)
5437 bb->loop_father = NULL;
5438 current_loops = NULL;
5440 ira_destroy ();
5442 lra (ira_dump_file);
5443 /* ???!!! Move it before lra () when we use ira_reg_equiv in
5444 LRA. */
5445 vec_free (reg_equivs);
5446 reg_equivs = NULL;
5447 need_dce = false;
5449 else
5451 df_set_flags (DF_NO_INSN_RESCAN);
5452 build_insn_chain ();
5454 need_dce = reload (get_insns (), ira_conflicts_p);
5458 timevar_pop (TV_RELOAD);
5460 timevar_push (TV_IRA);
5462 if (ira_conflicts_p && ! ira_use_lra_p)
5464 ira_free (ira_spilled_reg_stack_slots);
5465 ira_finish_assign ();
5468 if (internal_flag_ira_verbose > 0 && ira_dump_file != NULL
5469 && overall_cost_before != ira_overall_cost)
5470 fprintf (ira_dump_file, "+++Overall after reload %" PRId64 "\n",
5471 ira_overall_cost);
5473 flag_ira_share_spill_slots = saved_flag_ira_share_spill_slots;
5475 if (! ira_use_lra_p)
5477 ira_destroy ();
5478 if (current_loops != NULL)
5480 loop_optimizer_finalize ();
5481 free_dominance_info (CDI_DOMINATORS);
5483 FOR_ALL_BB_FN (bb, cfun)
5484 bb->loop_father = NULL;
5485 current_loops = NULL;
5487 regstat_free_ri ();
5488 regstat_free_n_sets_and_refs ();
5491 if (optimize)
5492 cleanup_cfg (CLEANUP_EXPENSIVE);
5494 finish_reg_equiv ();
5496 bitmap_obstack_release (&ira_bitmap_obstack);
5497 #ifndef IRA_NO_OBSTACK
5498 obstack_free (&ira_obstack, NULL);
5499 #endif
5501 /* The code after the reload has changed so much that at this point
5502 we might as well just rescan everything. Note that
5503 df_rescan_all_insns is not going to help here because it does not
5504 touch the artificial uses and defs. */
5505 df_finish_pass (true);
5506 df_scan_alloc (NULL);
5507 df_scan_blocks ();
5509 if (optimize > 1)
5511 df_live_add_problem ();
5512 df_live_set_all_dirty ();
5515 if (optimize)
5516 df_analyze ();
5518 if (need_dce && optimize)
5519 run_fast_dce ();
5521 /* Diagnose uses of the hard frame pointer when it is used as a global
5522 register. Often we can get away with letting the user appropriate
5523 the frame pointer, but we should let them know when code generation
5524 makes that impossible. */
5525 if (global_regs[HARD_FRAME_POINTER_REGNUM] && frame_pointer_needed)
5527 tree decl = global_regs_decl[HARD_FRAME_POINTER_REGNUM];
5528 error_at (DECL_SOURCE_LOCATION (current_function_decl),
5529 "frame pointer required, but reserved");
5530 inform (DECL_SOURCE_LOCATION (decl), "for %qD", decl);
5533 if (pic_offset_table_regno != INVALID_REGNUM)
5534 pic_offset_table_rtx = gen_rtx_REG (Pmode, pic_offset_table_regno);
5536 timevar_pop (TV_IRA);
5539 /* Run the integrated register allocator. */
5541 namespace {
5543 const pass_data pass_data_ira =
5545 RTL_PASS, /* type */
5546 "ira", /* name */
5547 OPTGROUP_NONE, /* optinfo_flags */
5548 TV_IRA, /* tv_id */
5549 0, /* properties_required */
5550 0, /* properties_provided */
5551 0, /* properties_destroyed */
5552 0, /* todo_flags_start */
5553 TODO_do_not_ggc_collect, /* todo_flags_finish */
5556 class pass_ira : public rtl_opt_pass
5558 public:
5559 pass_ira (gcc::context *ctxt)
5560 : rtl_opt_pass (pass_data_ira, ctxt)
5563 /* opt_pass methods: */
5564 virtual bool gate (function *)
5566 return !targetm.no_register_allocation;
5568 virtual unsigned int execute (function *)
5570 ira (dump_file);
5571 return 0;
5574 }; // class pass_ira
5576 } // anon namespace
5578 rtl_opt_pass *
5579 make_pass_ira (gcc::context *ctxt)
5581 return new pass_ira (ctxt);
5584 namespace {
5586 const pass_data pass_data_reload =
5588 RTL_PASS, /* type */
5589 "reload", /* name */
5590 OPTGROUP_NONE, /* optinfo_flags */
5591 TV_RELOAD, /* tv_id */
5592 0, /* properties_required */
5593 0, /* properties_provided */
5594 0, /* properties_destroyed */
5595 0, /* todo_flags_start */
5596 0, /* todo_flags_finish */
5599 class pass_reload : public rtl_opt_pass
5601 public:
5602 pass_reload (gcc::context *ctxt)
5603 : rtl_opt_pass (pass_data_reload, ctxt)
5606 /* opt_pass methods: */
5607 virtual bool gate (function *)
5609 return !targetm.no_register_allocation;
5611 virtual unsigned int execute (function *)
5613 do_reload ();
5614 return 0;
5617 }; // class pass_reload
5619 } // anon namespace
5621 rtl_opt_pass *
5622 make_pass_reload (gcc::context *ctxt)
5624 return new pass_reload (ctxt);