Google Git
Sign in
gnu / gcc / 19220ca6aa79921cc431e41f25986e16410c7a6a / . / gcc / gcse.c
blob: fdd2767e14331085e560a9bbd0b65268513681cb [file] [log] [blame]
/* Global common subexpression elimination/Partial redundancy elimination
and global constant/copy propagation for GNU compiler.
Copyright (C) 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004
Free Software Foundation, Inc.
This file is part of GCC.
GCC is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 2, or (at your option) any later
version.
GCC is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING. If not, write to the Free
Software Foundation, 59 Temple Place - Suite 330, Boston, MA
02111-1307, USA. */
/* TODO
- reordering of memory allocation and freeing to be more space efficient
- do rough calc of how many regs are needed in each block, and a rough
calc of how many regs are available in each class and use that to
throttle back the code in cases where RTX_COST is minimal.
- a store to the same address as a load does not kill the load if the
source of the store is also the destination of the load. Handling this
allows more load motion, particularly out of loops.
- ability to realloc sbitmap vectors would allow one initial computation
of reg_set_in_block with only subsequent additions, rather than
recomputing it for each pass
*/
/* References searched while implementing this.
Compilers Principles, Techniques and Tools
Aho, Sethi, Ullman
Addison-Wesley, 1988
Global Optimization by Suppression of Partial Redundancies
E. Morel, C. Renvoise
communications of the acm, Vol. 22, Num. 2, Feb. 1979
A Portable Machine-Independent Global Optimizer - Design and Measurements
Frederick Chow
Stanford Ph.D. thesis, Dec. 1983
A Fast Algorithm for Code Movement Optimization
D.M. Dhamdhere
SIGPLAN Notices, Vol. 23, Num. 10, Oct. 1988
A Solution to a Problem with Morel and Renvoise's
Global Optimization by Suppression of Partial Redundancies
K-H Drechsler, M.P. Stadel
ACM TOPLAS, Vol. 10, Num. 4, Oct. 1988
Practical Adaptation of the Global Optimization
Algorithm of Morel and Renvoise
D.M. Dhamdhere
ACM TOPLAS, Vol. 13, Num. 2. Apr. 1991
Efficiently Computing Static Single Assignment Form and the Control
Dependence Graph
R. Cytron, J. Ferrante, B.K. Rosen, M.N. Wegman, and F.K. Zadeck
ACM TOPLAS, Vol. 13, Num. 4, Oct. 1991
Lazy Code Motion
J. Knoop, O. Ruthing, B. Steffen
ACM SIGPLAN Notices Vol. 27, Num. 7, Jul. 1992, '92 Conference on PLDI
What's In a Region? Or Computing Control Dependence Regions in Near-Linear
Time for Reducible Flow Control
Thomas Ball
ACM Letters on Programming Languages and Systems,
Vol. 2, Num. 1-4, Mar-Dec 1993
An Efficient Representation for Sparse Sets
Preston Briggs, Linda Torczon
ACM Letters on Programming Languages and Systems,
Vol. 2, Num. 1-4, Mar-Dec 1993
A Variation of Knoop, Ruthing, and Steffen's Lazy Code Motion
K-H Drechsler, M.P. Stadel
ACM SIGPLAN Notices, Vol. 28, Num. 5, May 1993
Partial Dead Code Elimination
J. Knoop, O. Ruthing, B. Steffen
ACM SIGPLAN Notices, Vol. 29, Num. 6, Jun. 1994
Effective Partial Redundancy Elimination
P. Briggs, K.D. Cooper
ACM SIGPLAN Notices, Vol. 29, Num. 6, Jun. 1994
The Program Structure Tree: Computing Control Regions in Linear Time
R. Johnson, D. Pearson, K. Pingali
ACM SIGPLAN Notices, Vol. 29, Num. 6, Jun. 1994
Optimal Code Motion: Theory and Practice
J. Knoop, O. Ruthing, B. Steffen
ACM TOPLAS, Vol. 16, Num. 4, Jul. 1994
The power of assignment motion
J. Knoop, O. Ruthing, B. Steffen
ACM SIGPLAN Notices Vol. 30, Num. 6, Jun. 1995, '95 Conference on PLDI
Global code motion / global value numbering
C. Click
ACM SIGPLAN Notices Vol. 30, Num. 6, Jun. 1995, '95 Conference on PLDI
Value Driven Redundancy Elimination
L.T. Simpson
Rice University Ph.D. thesis, Apr. 1996
Value Numbering
L.T. Simpson
Massively Scalar Compiler Project, Rice University, Sep. 1996
High Performance Compilers for Parallel Computing
Michael Wolfe
Addison-Wesley, 1996
Advanced Compiler Design and Implementation
Steven Muchnick
Morgan Kaufmann, 1997
Building an Optimizing Compiler
Robert Morgan
Digital Press, 1998
People wishing to speed up the code here should read:
Elimination Algorithms for Data Flow Analysis
B.G. Ryder, M.C. Paull
ACM Computing Surveys, Vol. 18, Num. 3, Sep. 1986
How to Analyze Large Programs Efficiently and Informatively
D.M. Dhamdhere, B.K. Rosen, F.K. Zadeck
ACM SIGPLAN Notices Vol. 27, Num. 7, Jul. 1992, '92 Conference on PLDI
People wishing to do something different can find various possibilities
in the above papers and elsewhere.
*/
#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "tm.h"
#include "toplev.h"
#include "rtl.h"
#include "tree.h"
#include "tm_p.h"
#include "regs.h"
#include "hard-reg-set.h"
#include "flags.h"
#include "real.h"
#include "insn-config.h"
#include "recog.h"
#include "basic-block.h"
#include "output.h"
#include "function.h"
#include "expr.h"
#include "except.h"
#include "ggc.h"
#include "params.h"
#include "cselib.h"
#include "intl.h"
#include "obstack.h"
/* Propagate flow information through back edges and thus enable PRE's
moving loop invariant calculations out of loops.
Originally this tended to create worse overall code, but several
improvements during the development of PRE seem to have made following
back edges generally a win.
Note much of the loop invariant code motion done here would normally
be done by loop.c, which has more heuristics for when to move invariants
out of loops. At some point we might need to move some of those
heuristics into gcse.c. */
/* We support GCSE via Partial Redundancy Elimination. PRE optimizations
are a superset of those done by GCSE.
We perform the following steps:
1) Compute basic block information.
2) Compute table of places where registers are set.
3) Perform copy/constant propagation.
4) Perform global cse.
5) Perform another pass of copy/constant propagation.
Two passes of copy/constant propagation are done because the first one
enables more GCSE and the second one helps to clean up the copies that
GCSE creates. This is needed more for PRE than for Classic because Classic
GCSE will try to use an existing register containing the common
subexpression rather than create a new one. This is harder to do for PRE
because of the code motion (which Classic GCSE doesn't do).
Expressions we are interested in GCSE-ing are of the form
(set (pseudo-reg) (expression)).
Function want_to_gcse_p says what these are.
PRE handles moving invariant expressions out of loops (by treating them as
partially redundant).
Eventually it would be nice to replace cse.c/gcse.c with SSA (static single
assignment) based GVN (global value numbering). L. T. Simpson's paper
(Rice University) on value numbering is a useful reference for this.
**********************
We used to support multiple passes but there are diminishing returns in
doing so. The first pass usually makes 90% of the changes that are doable.
A second pass can make a few more changes made possible by the first pass.
Experiments show any further passes don't make enough changes to justify
the expense.
A study of spec92 using an unlimited number of passes:
[1 pass] = 1208 substitutions, [2] = 577, [3] = 202, [4] = 192, [5] = 83,
[6] = 34, [7] = 17, [8] = 9, [9] = 4, [10] = 4, [11] = 2,
[12] = 2, [13] = 1, [15] = 1, [16] = 2, [41] = 1
It was found doing copy propagation between each pass enables further
substitutions.
PRE is quite expensive in complicated functions because the DFA can take
awhile to converge. Hence we only perform one pass. The parameter max-gcse-passes can
be modified if one wants to experiment.
**********************
The steps for PRE are:
1) Build the hash table of expressions we wish to GCSE (expr_hash_table).
2) Perform the data flow analysis for PRE.
3) Delete the redundant instructions
4) Insert the required copies [if any] that make the partially
redundant instructions fully redundant.
5) For other reaching expressions, insert an instruction to copy the value
to a newly created pseudo that will reach the redundant instruction.
The deletion is done first so that when we do insertions we
know which pseudo reg to use.
Various papers have argued that PRE DFA is expensive (O(n^2)) and others
argue it is not. The number of iterations for the algorithm to converge
is typically 2-4 so I don't view it as that expensive (relatively speaking).
PRE GCSE depends heavily on the second CSE pass to clean up the copies
we create. To make an expression reach the place where it's redundant,
the result of the expression is copied to a new register, and the redundant
expression is deleted by replacing it with this new register. Classic GCSE
doesn't have this problem as much as it computes the reaching defs of
each register in each block and thus can try to use an existing register.
**********************
A fair bit of simplicity is created by creating small functions for simple
tasks, even when the function is only called in one place. This may
measurably slow things down [or may not] by creating more function call
overhead than is necessary. The source is laid out so that it's trivial
to make the affected functions inline so that one can measure what speed
up, if any, can be achieved, and maybe later when things settle things can
be rearranged.
Help stamp out big monolithic functions! */
/* GCSE global vars. */
/* -dG dump file. */
static FILE *gcse_file;
/* Note whether or not we should run jump optimization after gcse. We
want to do this for two cases.
* If we changed any jumps via cprop.
* If we added any labels via edge splitting. */
static int run_jump_opt_after_gcse;
/* Bitmaps are normally not included in debugging dumps.
However it's useful to be able to print them from GDB.
We could create special functions for this, but it's simpler to
just allow passing stderr to the dump_foo fns. Since stderr can
be a macro, we store a copy here. */
static FILE *debug_stderr;
/* An obstack for our working variables. */
static struct obstack gcse_obstack;
struct reg_use {rtx reg_rtx; };
/* Hash table of expressions. */
struct expr
{
/* The expression (SET_SRC for expressions, PATTERN for assignments). */
rtx expr;
/* Index in the available expression bitmaps. */
int bitmap_index;
/* Next entry with the same hash. */
struct expr *next_same_hash;
/* List of anticipatable occurrences in basic blocks in the function.
An "anticipatable occurrence" is one that is the first occurrence in the
basic block, the operands are not modified in the basic block prior
to the occurrence and the output is not used between the start of
the block and the occurrence. */
struct occr *antic_occr;
/* List of available occurrence in basic blocks in the function.
An "available occurrence" is one that is the last occurrence in the
basic block and the operands are not modified by following statements in
the basic block [including this insn]. */
struct occr *avail_occr;
/* Non-null if the computation is PRE redundant.
The value is the newly created pseudo-reg to record a copy of the
expression in all the places that reach the redundant copy. */
rtx reaching_reg;
};
/* Occurrence of an expression.
There is one per basic block. If a pattern appears more than once the
last appearance is used [or first for anticipatable expressions]. */
struct occr
{
/* Next occurrence of this expression. */
struct occr *next;
/* The insn that computes the expression. */
rtx insn;
/* Nonzero if this [anticipatable] occurrence has been deleted. */
char deleted_p;
/* Nonzero if this [available] occurrence has been copied to
reaching_reg. */
/* ??? This is mutually exclusive with deleted_p, so they could share
the same byte. */
char copied_p;
};
/* Expression and copy propagation hash tables.
Each hash table is an array of buckets.
??? It is known that if it were an array of entries, structure elements
`next_same_hash' and `bitmap_index' wouldn't be necessary. However, it is
not clear whether in the final analysis a sufficient amount of memory would
be saved as the size of the available expression bitmaps would be larger
[one could build a mapping table without holes afterwards though].
Someday I'll perform the computation and figure it out. */
struct hash_table
{
/* The table itself.
This is an array of `expr_hash_table_size' elements. */
struct expr **table;
/* Size of the hash table, in elements. */
unsigned int size;
/* Number of hash table elements. */
unsigned int n_elems;
/* Whether the table is expression of copy propagation one. */
int set_p;
};
/* Expression hash table. */
static struct hash_table expr_hash_table;
/* Copy propagation hash table. */
static struct hash_table set_hash_table;
/* Mapping of uids to cuids.
Only real insns get cuids. */
static int *uid_cuid;
/* Highest UID in UID_CUID. */
static int max_uid;
/* Get the cuid of an insn. */
#ifdef ENABLE_CHECKING
#define INSN_CUID(INSN) (INSN_UID (INSN) > max_uid ? (abort (), 0) : uid_cuid[INSN_UID (INSN)])
#else
#define INSN_CUID(INSN) (uid_cuid[INSN_UID (INSN)])
#endif
/* Number of cuids. */
static int max_cuid;
/* Mapping of cuids to insns. */
static rtx *cuid_insn;
/* Get insn from cuid. */
#define CUID_INSN(CUID) (cuid_insn[CUID])
/* Maximum register number in function prior to doing gcse + 1.
Registers created during this pass have regno >= max_gcse_regno.
This is named with "gcse" to not collide with global of same name. */
static unsigned int max_gcse_regno;
/* Table of registers that are modified.
For each register, each element is a list of places where the pseudo-reg
is set.
For simplicity, GCSE is done on sets of pseudo-regs only. PRE GCSE only
requires knowledge of which blocks kill which regs [and thus could use
a bitmap instead of the lists `reg_set_table' uses].
`reg_set_table' and could be turned into an array of bitmaps (num-bbs x
num-regs) [however perhaps it may be useful to keep the data as is]. One
advantage of recording things this way is that `reg_set_table' is fairly
sparse with respect to pseudo regs but for hard regs could be fairly dense
[relatively speaking]. And recording sets of pseudo-regs in lists speeds
up functions like compute_transp since in the case of pseudo-regs we only
need to iterate over the number of times a pseudo-reg is set, not over the
number of basic blocks [clearly there is a bit of a slow down in the cases
where a pseudo is set more than once in a block, however it is believed
that the net effect is to speed things up]. This isn't done for hard-regs
because recording call-clobbered hard-regs in `reg_set_table' at each
function call can consume a fair bit of memory, and iterating over
hard-regs stored this way in compute_transp will be more expensive. */
typedef struct reg_set
{
/* The next setting of this register. */
struct reg_set *next;
/* The insn where it was set. */
rtx insn;
} reg_set;
static reg_set **reg_set_table;
/* Size of `reg_set_table'.
The table starts out at max_gcse_regno + slop, and is enlarged as
necessary. */
static int reg_set_table_size;
/* Amount to grow `reg_set_table' by when it's full. */
#define REG_SET_TABLE_SLOP 100
/* This is a list of expressions which are MEMs and will be used by load
or store motion.
Load motion tracks MEMs which aren't killed by
anything except itself. (ie, loads and stores to a single location).
We can then allow movement of these MEM refs with a little special
allowance. (all stores copy the same value to the reaching reg used
for the loads). This means all values used to store into memory must have
no side effects so we can re-issue the setter value.
Store Motion uses this structure as an expression table to track stores
which look interesting, and might be moveable towards the exit block. */
struct ls_expr
{
struct expr * expr; /* Gcse expression reference for LM. */
rtx pattern; /* Pattern of this mem. */
rtx pattern_regs; /* List of registers mentioned by the mem. */
rtx loads; /* INSN list of loads seen. */
rtx stores; /* INSN list of stores seen. */
struct ls_expr * next; /* Next in the list. */
int invalid; /* Invalid for some reason. */
int index; /* If it maps to a bitmap index. */
unsigned int hash_index; /* Index when in a hash table. */
rtx reaching_reg; /* Register to use when re-writing. */
};
/* Array of implicit set patterns indexed by basic block index. */
static rtx *implicit_sets;
/* Head of the list of load/store memory refs. */
static struct ls_expr * pre_ldst_mems = NULL;
/* Bitmap containing one bit for each register in the program.
Used when performing GCSE to track which registers have been set since
the start of the basic block. */
static regset reg_set_bitmap;
/* For each block, a bitmap of registers set in the block.
This is used by expr_killed_p and compute_transp.
It is computed during hash table computation and not by compute_sets
as it includes registers added since the last pass (or between cprop and
gcse) and it's currently not easy to realloc sbitmap vectors. */
static sbitmap *reg_set_in_block;
/* Array, indexed by basic block number for a list of insns which modify
memory within that block. */
static rtx * modify_mem_list;
bitmap modify_mem_list_set;
/* This array parallels modify_mem_list, but is kept canonicalized. */
static rtx * canon_modify_mem_list;
bitmap canon_modify_mem_list_set;
/* Various variables for statistics gathering. */
/* Memory used in a pass.
This isn't intended to be absolutely precise. Its intent is only
to keep an eye on memory usage. */
static int bytes_used;
/* GCSE substitutions made. */
static int gcse_subst_count;
/* Number of copy instructions created. */
static int gcse_create_count;
/* Number of constants propagated. */
static int const_prop_count;
/* Number of copys propagated. */
static int copy_prop_count;
/* These variables are used by classic GCSE.
Normally they'd be defined a bit later, but `rd_gen' needs to
be declared sooner. */
/* Each block has a bitmap of each type.
The length of each blocks bitmap is:
max_cuid - for reaching definitions
n_exprs - for available expressions
Thus we view the bitmaps as 2 dimensional arrays. i.e.
rd_kill[block_num][cuid_num]
ae_kill[block_num][expr_num] */
/* For reaching defs */
static sbitmap *rd_kill, *rd_gen, *reaching_defs, *rd_out;
/* for available exprs */
static sbitmap *ae_kill, *ae_gen, *ae_in, *ae_out;
/* Objects of this type are passed around by the null-pointer check
removal routines. */
struct null_pointer_info
{
/* The basic block being processed. */
basic_block current_block;
/* The first register to be handled in this pass. */
unsigned int min_reg;
/* One greater than the last register to be handled in this pass. */
unsigned int max_reg;
sbitmap *nonnull_local;
sbitmap *nonnull_killed;
};
static void compute_can_copy (void);
static void *gmalloc (size_t) ATTRIBUTE_MALLOC;
static void *gcalloc (size_t, size_t) ATTRIBUTE_MALLOC;
static void *grealloc (void *, size_t);
static void *gcse_alloc (unsigned long);
static void alloc_gcse_mem (rtx);
static void free_gcse_mem (void);
static void alloc_reg_set_mem (int);
static void free_reg_set_mem (void);
static int get_bitmap_width (int, int, int);
static void record_one_set (int, rtx);
static void replace_one_set (int, rtx, rtx);
static void record_set_info (rtx, rtx, void *);
static void compute_sets (rtx);
static void hash_scan_insn (rtx, struct hash_table *, int);
static void hash_scan_set (rtx, rtx, struct hash_table *);
static void hash_scan_clobber (rtx, rtx, struct hash_table *);
static void hash_scan_call (rtx, rtx, struct hash_table *);
static int want_to_gcse_p (rtx);
static bool gcse_constant_p (rtx);
static int oprs_unchanged_p (rtx, rtx, int);
static int oprs_anticipatable_p (rtx, rtx);
static int oprs_available_p (rtx, rtx);
static void insert_expr_in_table (rtx, enum machine_mode, rtx, int, int,
struct hash_table *);
static void insert_set_in_table (rtx, rtx, struct hash_table *);
static unsigned int hash_expr (rtx, enum machine_mode, int *, int);
static unsigned int hash_expr_1 (rtx, enum machine_mode, int *);
static unsigned int hash_string_1 (const char *);
static unsigned int hash_set (int, int);
static int expr_equiv_p (rtx, rtx);
static void record_last_reg_set_info (rtx, int);
static void record_last_mem_set_info (rtx);
static void record_last_set_info (rtx, rtx, void *);
static void compute_hash_table (struct hash_table *);
static void alloc_hash_table (int, struct hash_table *, int);
static void free_hash_table (struct hash_table *);
static void compute_hash_table_work (struct hash_table *);
static void dump_hash_table (FILE *, const char *, struct hash_table *);
static struct expr *lookup_expr (rtx, struct hash_table *);
static struct expr *lookup_set (unsigned int, struct hash_table *);
static struct expr *next_set (unsigned int, struct expr *);
static void reset_opr_set_tables (void);
static int oprs_not_set_p (rtx, rtx);
static void mark_call (rtx);
static void mark_set (rtx, rtx);
static void mark_clobber (rtx, rtx);
static void mark_oprs_set (rtx);
static void alloc_cprop_mem (int, int);
static void free_cprop_mem (void);
static void compute_transp (rtx, int, sbitmap *, int);
static void compute_transpout (void);
static void compute_local_properties (sbitmap *, sbitmap *, sbitmap *,
struct hash_table *);
static void compute_cprop_data (void);
static void find_used_regs (rtx *, void *);
static int try_replace_reg (rtx, rtx, rtx);
static struct expr *find_avail_set (int, rtx);
static int cprop_jump (basic_block, rtx, rtx, rtx, rtx);
static void mems_conflict_for_gcse_p (rtx, rtx, void *);
static int load_killed_in_block_p (basic_block, int, rtx, int);
static void canon_list_insert (rtx, rtx, void *);
static int cprop_insn (rtx, int);
static int cprop (int);
static void find_implicit_sets (void);
static int one_cprop_pass (int, int, int);
static bool constprop_register (rtx, rtx, rtx, int);
static struct expr *find_bypass_set (int, int);
static bool reg_killed_on_edge (rtx, edge);
static int bypass_block (basic_block, rtx, rtx);
static int bypass_conditional_jumps (void);
static void alloc_pre_mem (int, int);
static void free_pre_mem (void);
static void compute_pre_data (void);
static int pre_expr_reaches_here_p (basic_block, struct expr *,
basic_block);
static void insert_insn_end_bb (struct expr *, basic_block, int);
static void pre_insert_copy_insn (struct expr *, rtx);
static void pre_insert_copies (void);
static int pre_delete (void);
static int pre_gcse (void);
static int one_pre_gcse_pass (int);
static void add_label_notes (rtx, rtx);
static void alloc_code_hoist_mem (int, int);
static void free_code_hoist_mem (void);
static void compute_code_hoist_vbeinout (void);
static void compute_code_hoist_data (void);
static int hoist_expr_reaches_here_p (basic_block, int, basic_block, char *);
static void hoist_code (void);
static int one_code_hoisting_pass (void);
static void alloc_rd_mem (int, int);
static void free_rd_mem (void);
static void handle_rd_kill_set (rtx, int, basic_block);
static void compute_kill_rd (void);
static void compute_rd (void);
static void alloc_avail_expr_mem (int, int);
static void free_avail_expr_mem (void);
static void compute_ae_gen (struct hash_table *);
static int expr_killed_p (rtx, basic_block);
static void compute_ae_kill (sbitmap *, sbitmap *, struct hash_table *);
static int expr_reaches_here_p (struct occr *, struct expr *, basic_block,
int);
static rtx computing_insn (struct expr *, rtx);
static int def_reaches_here_p (rtx, rtx);
static int can_disregard_other_sets (struct reg_set **, rtx, int);
static int handle_avail_expr (rtx, struct expr *);
static int classic_gcse (void);
static int one_classic_gcse_pass (int);
static void invalidate_nonnull_info (rtx, rtx, void *);
static int delete_null_pointer_checks_1 (unsigned int *, sbitmap *, sbitmap *,
struct null_pointer_info *);
static rtx process_insert_insn (struct expr *);
static int pre_edge_insert (struct edge_list *, struct expr **);
static int expr_reaches_here_p_work (struct occr *, struct expr *,
basic_block, int, char *);
static int pre_expr_reaches_here_p_work (basic_block, struct expr *,
basic_block, char *);
static struct ls_expr * ldst_entry (rtx);
static void free_ldst_entry (struct ls_expr *);
static void free_ldst_mems (void);
static void print_ldst_list (FILE *);
static struct ls_expr * find_rtx_in_ldst (rtx);
static int enumerate_ldsts (void);
static inline struct ls_expr * first_ls_expr (void);
static inline struct ls_expr * next_ls_expr (struct ls_expr *);
static int simple_mem (rtx);
static void invalidate_any_buried_refs (rtx);
static void compute_ld_motion_mems (void);
static void trim_ld_motion_mems (void);
static void update_ld_motion_stores (struct expr *);
static void reg_set_info (rtx, rtx, void *);
static void reg_clear_last_set (rtx, rtx, void *);
static bool store_ops_ok (rtx, int *);
static rtx extract_mentioned_regs (rtx);
static rtx extract_mentioned_regs_helper (rtx, rtx);
static void find_moveable_store (rtx, int *, int *);
static int compute_store_table (void);
static bool load_kills_store (rtx, rtx, int);
static bool find_loads (rtx, rtx, int);
static bool store_killed_in_insn (rtx, rtx, rtx, int);
static bool store_killed_after (rtx, rtx, rtx, basic_block, int *, rtx *);
static bool store_killed_before (rtx, rtx, rtx, basic_block, int *);
static void build_store_vectors (void);
static void insert_insn_start_bb (rtx, basic_block);
static int insert_store (struct ls_expr *, edge);
static void remove_reachable_equiv_notes (basic_block, struct ls_expr *);
static void replace_store_insn (rtx, rtx, basic_block, struct ls_expr *);
static void delete_store (struct ls_expr *, basic_block);
static void free_store_memory (void);
static void store_motion (void);
static void free_insn_expr_list_list (rtx *);
static void clear_modify_mem_tables (void);
static void free_modify_mem_tables (void);
static rtx gcse_emit_move_after (rtx, rtx, rtx);
static void local_cprop_find_used_regs (rtx *, void *);
static bool do_local_cprop (rtx, rtx, int, rtx*);
static bool adjust_libcall_notes (rtx, rtx, rtx, rtx*);
static void local_cprop_pass (int);
static bool is_too_expensive (const char *);
/* Entry point for global common subexpression elimination.
F is the first instruction in the function. */
int
gcse_main (rtx f, FILE *file)
{
int changed, pass;
/* Bytes used at start of pass. */
int initial_bytes_used;
/* Maximum number of bytes used by a pass. */
int max_pass_bytes;
/* Point to release obstack data from for each pass. */
char *gcse_obstack_bottom;
/* We do not construct an accurate cfg in functions which call
setjmp, so just punt to be safe. */
if (current_function_calls_setjmp)
return 0;
/* Assume that we do not need to run jump optimizations after gcse. */
run_jump_opt_after_gcse = 0;
/* For calling dump_foo fns from gdb. */
debug_stderr = stderr;
gcse_file = file;
/* Identify the basic block information for this function, including
successors and predecessors. */
max_gcse_regno = max_reg_num ();
if (file)
dump_flow_info (file);
/* Return if there's nothing to do, or it is too expensive. */
if (n_basic_blocks <= 1 || is_too_expensive (_("GCSE disabled")))
return 0;
gcc_obstack_init (&gcse_obstack);
bytes_used = 0;
/* We need alias. */
init_alias_analysis ();
/* Record where pseudo-registers are set. This data is kept accurate
during each pass. ??? We could also record hard-reg information here
[since it's unchanging], however it is currently done during hash table
computation.
It may be tempting to compute MEM set information here too, but MEM sets
will be subject to code motion one day and thus we need to compute
information about memory sets when we build the hash tables. */
alloc_reg_set_mem (max_gcse_regno);
compute_sets (f);
pass = 0;
initial_bytes_used = bytes_used;
max_pass_bytes = 0;
gcse_obstack_bottom = gcse_alloc (1);
changed = 1;
while (changed && pass < MAX_GCSE_PASSES)
{
changed = 0;
if (file)
fprintf (file, "GCSE pass %d\n\n", pass + 1);
/* Initialize bytes_used to the space for the pred/succ lists,
and the reg_set_table data. */
bytes_used = initial_bytes_used;
/* Each pass may create new registers, so recalculate each time. */
max_gcse_regno = max_reg_num ();
alloc_gcse_mem (f);
/* Don't allow constant propagation to modify jumps
during this pass. */
changed = one_cprop_pass (pass + 1, 0, 0);
if (optimize_size)
changed |= one_classic_gcse_pass (pass + 1);
else
{
changed |= one_pre_gcse_pass (pass + 1);
/* We may have just created new basic blocks. Release and
recompute various things which are sized on the number of
basic blocks. */
if (changed)
{
free_modify_mem_tables ();
modify_mem_list = gcalloc (last_basic_block, sizeof (rtx));
canon_modify_mem_list = gcalloc (last_basic_block, sizeof (rtx));
}
free_reg_set_mem ();
alloc_reg_set_mem (max_reg_num ());
compute_sets (f);
run_jump_opt_after_gcse = 1;
}
if (max_pass_bytes < bytes_used)
max_pass_bytes = bytes_used;
/* Free up memory, then reallocate for code hoisting. We can
not re-use the existing allocated memory because the tables
will not have info for the insns or registers created by
partial redundancy elimination. */
free_gcse_mem ();
/* It does not make sense to run code hoisting unless we are optimizing
for code size -- it rarely makes programs faster, and can make
them bigger if we did partial redundancy elimination (when optimizing
for space, we use a classic gcse algorithm instead of partial
redundancy algorithms). */
if (optimize_size)
{
max_gcse_regno = max_reg_num ();
alloc_gcse_mem (f);
changed |= one_code_hoisting_pass ();
free_gcse_mem ();
if (max_pass_bytes < bytes_used)
max_pass_bytes = bytes_used;
}
if (file)
{
fprintf (file, "\n");
fflush (file);
}
obstack_free (&gcse_obstack, gcse_obstack_bottom);
pass++;
}
/* Do one last pass of copy propagation, including cprop into
conditional jumps. */
max_gcse_regno = max_reg_num ();
alloc_gcse_mem (f);
/* This time, go ahead and allow cprop to alter jumps. */
one_cprop_pass (pass + 1, 1, 0);
free_gcse_mem ();
if (file)
{
fprintf (file, "GCSE of %s: %d basic blocks, ",
current_function_name (), n_basic_blocks);
fprintf (file, "%d pass%s, %d bytes\n\n",
pass, pass > 1 ? "es" : "", max_pass_bytes);
}
obstack_free (&gcse_obstack, NULL);
free_reg_set_mem ();
/* We are finished with alias. */
end_alias_analysis ();
allocate_reg_info (max_reg_num (), FALSE, FALSE);
if (!optimize_size && flag_gcse_sm)
store_motion ();
/* Record where pseudo-registers are set. */
return run_jump_opt_after_gcse;
}
/* Misc. utilities. */
/* Nonzero for each mode that supports (set (reg) (reg)).
This is trivially true for integer and floating point values.
It may or may not be true for condition codes. */
static char can_copy[(int) NUM_MACHINE_MODES];
/* Compute which modes support reg/reg copy operations. */
static void
compute_can_copy (void)
{
int i;
#ifndef AVOID_CCMODE_COPIES
rtx reg, insn;
#endif
memset (can_copy, 0, NUM_MACHINE_MODES);
start_sequence ();
for (i = 0; i < NUM_MACHINE_MODES; i++)
if (GET_MODE_CLASS (i) == MODE_CC)
{
#ifdef AVOID_CCMODE_COPIES
can_copy[i] = 0;
#else
reg = gen_rtx_REG ((enum machine_mode) i, LAST_VIRTUAL_REGISTER + 1);
insn = emit_insn (gen_rtx_SET (VOIDmode, reg, reg));
if (recog (PATTERN (insn), insn, NULL) >= 0)
can_copy[i] = 1;
#endif
}
else
can_copy[i] = 1;
end_sequence ();
}
/* Returns whether the mode supports reg/reg copy operations. */
bool
can_copy_p (enum machine_mode mode)
{
static bool can_copy_init_p = false;
if (! can_copy_init_p)
{
compute_can_copy ();
can_copy_init_p = true;
}
return can_copy[mode] != 0;
}
/* Cover function to xmalloc to record bytes allocated. */
static void *
gmalloc (size_t size)
{
bytes_used += size;
return xmalloc (size);
}
/* Cover function to xcalloc to record bytes allocated. */
static void *
gcalloc (size_t nelem, size_t elsize)
{
bytes_used += nelem * elsize;
return xcalloc (nelem, elsize);
}
/* Cover function to xrealloc.
We don't record the additional size since we don't know it.
It won't affect memory usage stats much anyway. */
static void *
grealloc (void *ptr, size_t size)
{
return xrealloc (ptr, size);
}
/* Cover function to obstack_alloc. */
static void *
gcse_alloc (unsigned long size)
{
bytes_used += size;
return obstack_alloc (&gcse_obstack, size);
}
/* Allocate memory for the cuid mapping array,
and reg/memory set tracking tables.
This is called at the start of each pass. */
static void
alloc_gcse_mem (rtx f)
{
int i;
rtx insn;
/* Find the largest UID and create a mapping from UIDs to CUIDs.
CUIDs are like UIDs except they increase monotonically, have no gaps,
and only apply to real insns. */
max_uid = get_max_uid ();
uid_cuid = gcalloc (max_uid + 1, sizeof (int));
for (insn = f, i = 0; insn; insn = NEXT_INSN (insn))
{
if (INSN_P (insn))
uid_cuid[INSN_UID (insn)] = i++;
else
uid_cuid[INSN_UID (insn)] = i;
}
/* Create a table mapping cuids to insns. */
max_cuid = i;
cuid_insn = gcalloc (max_cuid + 1, sizeof (rtx));
for (insn = f, i = 0; insn; insn = NEXT_INSN (insn))
if (INSN_P (insn))
CUID_INSN (i++) = insn;
/* Allocate vars to track sets of regs. */
reg_set_bitmap = BITMAP_XMALLOC ();
/* Allocate vars to track sets of regs, memory per block. */
reg_set_in_block = sbitmap_vector_alloc (last_basic_block, max_gcse_regno);
/* Allocate array to keep a list of insns which modify memory in each
basic block. */
modify_mem_list = gcalloc (last_basic_block, sizeof (rtx));
canon_modify_mem_list = gcalloc (last_basic_block, sizeof (rtx));
modify_mem_list_set = BITMAP_XMALLOC ();
canon_modify_mem_list_set = BITMAP_XMALLOC ();
}
/* Free memory allocated by alloc_gcse_mem. */
static void
free_gcse_mem (void)
{
free (uid_cuid);
free (cuid_insn);
BITMAP_XFREE (reg_set_bitmap);
sbitmap_vector_free (reg_set_in_block);
free_modify_mem_tables ();
BITMAP_XFREE (modify_mem_list_set);
BITMAP_XFREE (canon_modify_mem_list_set);
}
/* Many of the global optimization algorithms work by solving dataflow
equations for various expressions. Initially, some local value is
computed for each expression in each block. Then, the values across the
various blocks are combined (by following flow graph edges) to arrive at
global values. Conceptually, each set of equations is independent. We
may therefore solve all the equations in parallel, solve them one at a
time, or pick any intermediate approach.
When you're going to need N two-dimensional bitmaps, each X (say, the
number of blocks) by Y (say, the number of expressions), call this
function. It's not important what X and Y represent; only that Y
correspond to the things that can be done in parallel. This function will
return an appropriate chunking factor C; you should solve C sets of
equations in parallel. By going through this function, we can easily
trade space against time; by solving fewer equations in parallel we use
less space. */
static int
get_bitmap_width (int n, int x, int y)
{
/* It's not really worth figuring out *exactly* how much memory will
be used by a particular choice. The important thing is to get
something approximately right. */
size_t max_bitmap_memory = 10 * 1024 * 1024;
/* The number of bytes we'd use for a single column of minimum
width. */
size_t column_size = n * x * sizeof (SBITMAP_ELT_TYPE);
/* Often, it's reasonable just to solve all the equations in
parallel. */
if (column_size * SBITMAP_SET_SIZE (y) <= max_bitmap_memory)
return y;
/* Otherwise, pick the largest width we can, without going over the
limit. */
return SBITMAP_ELT_BITS * ((max_bitmap_memory + column_size - 1)
/ column_size);
}
/* Compute the local properties of each recorded expression.
Local properties are those that are defined by the block, irrespective of
other blocks.
An expression is transparent in a block if its operands are not modified
in the block.
An expression is computed (locally available) in a block if it is computed
at least once and expression would contain the same value if the
computation was moved to the end of the block.
An expression is locally anticipatable in a block if it is computed at
least once and expression would contain the same value if the computation
was moved to the beginning of the block.
We call this routine for cprop, pre and code hoisting. They all compute
basically the same information and thus can easily share this code.
TRANSP, COMP, and ANTLOC are destination sbitmaps for recording local
properties. If NULL, then it is not necessary to compute or record that
particular property.
TABLE controls which hash table to look at. If it is set hash table,
additionally, TRANSP is computed as ~TRANSP, since this is really cprop's
ABSALTERED. */
static void
compute_local_properties (sbitmap *transp, sbitmap *comp, sbitmap *antloc, struct hash_table *table)
{
unsigned int i;
/* Initialize any bitmaps that were passed in. */
if (transp)
{
if (table->set_p)
sbitmap_vector_zero (transp, last_basic_block);
else
sbitmap_vector_ones (transp, last_basic_block);
}
if (comp)
sbitmap_vector_zero (comp, last_basic_block);
if (antloc)
sbitmap_vector_zero (antloc, last_basic_block);
for (i = 0; i < table->size; i++)
{
struct expr *expr;
for (expr = table->table[i]; expr != NULL; expr = expr->next_same_hash)
{
int indx = expr->bitmap_index;
struct occr *occr;
/* The expression is transparent in this block if it is not killed.
We start by assuming all are transparent [none are killed], and
then reset the bits for those that are. */
if (transp)
compute_transp (expr->expr, indx, transp, table->set_p);
/* The occurrences recorded in antic_occr are exactly those that
we want to set to nonzero in ANTLOC. */
if (antloc)
for (occr = expr->antic_occr; occr != NULL; occr = occr->next)
{
SET_BIT (antloc[BLOCK_NUM (occr->insn)], indx);
/* While we're scanning the table, this is a good place to
initialize this. */
occr->deleted_p = 0;
}
/* The occurrences recorded in avail_occr are exactly those that
we want to set to nonzero in COMP. */
if (comp)
for (occr = expr->avail_occr; occr != NULL; occr = occr->next)
{
SET_BIT (comp[BLOCK_NUM (occr->insn)], indx);
/* While we're scanning the table, this is a good place to
initialize this. */
occr->copied_p = 0;
}
/* While we're scanning the table, this is a good place to
initialize this. */
expr->reaching_reg = 0;
}
}
}
/* Register set information.
`reg_set_table' records where each register is set or otherwise
modified. */
static struct obstack reg_set_obstack;
static void
alloc_reg_set_mem (int n_regs)
{
reg_set_table_size = n_regs + REG_SET_TABLE_SLOP;
reg_set_table = gcalloc (reg_set_table_size, sizeof (struct reg_set *));
gcc_obstack_init (&reg_set_obstack);
}
static void
free_reg_set_mem (void)
{
free (reg_set_table);
obstack_free (&reg_set_obstack, NULL);
}
/* An OLD_INSN that used to set REGNO was replaced by NEW_INSN.
Update the corresponding `reg_set_table' entry accordingly.
We assume that NEW_INSN is not already recorded in reg_set_table[regno]. */
static void
replace_one_set (int regno, rtx old_insn, rtx new_insn)
{
struct reg_set *reg_info;
if (regno >= reg_set_table_size)
return;
for (reg_info = reg_set_table[regno]; reg_info; reg_info = reg_info->next)
if (reg_info->insn == old_insn)
{
reg_info->insn = new_insn;
break;
}
}
/* Record REGNO in the reg_set table. */
static void
record_one_set (int regno, rtx insn)
{
/* Allocate a new reg_set element and link it onto the list. */
struct reg_set *new_reg_info;
/* If the table isn't big enough, enlarge it. */
if (regno >= reg_set_table_size)
{
int new_size = regno + REG_SET_TABLE_SLOP;
reg_set_table = grealloc (reg_set_table,
new_size * sizeof (struct reg_set *));
memset (reg_set_table + reg_set_table_size, 0,
(new_size - reg_set_table_size) * sizeof (struct reg_set *));
reg_set_table_size = new_size;
}
new_reg_info = obstack_alloc (&reg_set_obstack, sizeof (struct reg_set));
bytes_used += sizeof (struct reg_set);
new_reg_info->insn = insn;
new_reg_info->next = reg_set_table[regno];
reg_set_table[regno] = new_reg_info;
}
/* Called from compute_sets via note_stores to handle one SET or CLOBBER in
an insn. The DATA is really the instruction in which the SET is
occurring. */
static void
record_set_info (rtx dest, rtx setter ATTRIBUTE_UNUSED, void *data)
{
rtx record_set_insn = (rtx) data;
if (GET_CODE (dest) == REG && REGNO (dest) >= FIRST_PSEUDO_REGISTER)
record_one_set (REGNO (dest), record_set_insn);
}
/* Scan the function and record each set of each pseudo-register.
This is called once, at the start of the gcse pass. See the comments for
`reg_set_table' for further documentation. */
static void
compute_sets (rtx f)
{
rtx insn;
for (insn = f; insn != 0; insn = NEXT_INSN (insn))
if (INSN_P (insn))
note_stores (PATTERN (insn), record_set_info, insn);
}
/* Hash table support. */
struct reg_avail_info
{
basic_block last_bb;
int first_set;
int last_set;
};
static struct reg_avail_info *reg_avail_info;
static basic_block current_bb;
/* See whether X, the source of a set, is something we want to consider for
GCSE. */
static GTY(()) rtx test_insn;
static int
want_to_gcse_p (rtx x)
{
int num_clobbers = 0;
int icode;
switch (GET_CODE (x))
{
case REG:
case SUBREG:
case CONST_INT:
case CONST_DOUBLE:
case CONST_VECTOR:
case CALL:
case CONSTANT_P_RTX:
return 0;
default:
break;
}
/* If this is a valid operand, we are OK. If it's VOIDmode, we aren't. */
if (general_operand (x, GET_MODE (x)))
return 1;
else if (GET_MODE (x) == VOIDmode)
return 0;
/* Otherwise, check if we can make a valid insn from it. First initialize
our test insn if we haven't already. */
if (test_insn == 0)
{
test_insn
= make_insn_raw (gen_rtx_SET (VOIDmode,
gen_rtx_REG (word_mode,
FIRST_PSEUDO_REGISTER * 2),
const0_rtx));
NEXT_INSN (test_insn) = PREV_INSN (test_insn) = 0;
}
/* Now make an insn like the one we would make when GCSE'ing and see if
valid. */
PUT_MODE (SET_DEST (PATTERN (test_insn)), GET_MODE (x));
SET_SRC (PATTERN (test_insn)) = x;
return ((icode = recog (PATTERN (test_insn), test_insn, &num_clobbers)) >= 0
&& (num_clobbers == 0 || ! added_clobbers_hard_reg_p (icode)));
}
/* Return nonzero if the operands of expression X are unchanged from the
start of INSN's basic block up to but not including INSN (if AVAIL_P == 0),
or from INSN to the end of INSN's basic block (if AVAIL_P != 0). */
static int
oprs_unchanged_p (rtx x, rtx insn, int avail_p)
{
int i, j;
enum rtx_code code;
const char *fmt;
if (x == 0)
return 1;
code = GET_CODE (x);
switch (code)
{
case REG:
{
struct reg_avail_info *info = &reg_avail_info[REGNO (x)];
if (info->last_bb != current_bb)
return 1;
if (avail_p)
return info->last_set < INSN_CUID (insn);
else
return info->first_set >= INSN_CUID (insn);
}
case MEM:
if (load_killed_in_block_p (current_bb, INSN_CUID (insn),
x, avail_p))
return 0;
else
return oprs_unchanged_p (XEXP (x, 0), insn, avail_p);
case PRE_DEC:
case PRE_INC:
case POST_DEC:
case POST_INC:
case PRE_MODIFY:
case POST_MODIFY:
return 0;
case PC:
case CC0: /*FIXME*/
case CONST:
case CONST_INT:
case CONST_DOUBLE:
case CONST_VECTOR:
case SYMBOL_REF:
case LABEL_REF:
case ADDR_VEC:
case ADDR_DIFF_VEC:
return 1;
default:
break;
}
for (i = GET_RTX_LENGTH (code) - 1, fmt = GET_RTX_FORMAT (code); i >= 0; i--)
{
if (fmt[i] == 'e')
{
/* If we are about to do the last recursive call needed at this
level, change it into iteration. This function is called enough
to be worth it. */
if (i == 0)
return oprs_unchanged_p (XEXP (x, i), insn, avail_p);
else if (! oprs_unchanged_p (XEXP (x, i), insn, avail_p))
return 0;
}
else if (fmt[i] == 'E')
for (j = 0; j < XVECLEN (x, i); j++)
if (! oprs_unchanged_p (XVECEXP (x, i, j), insn, avail_p))
return 0;
}
return 1;
}
/* Used for communication between mems_conflict_for_gcse_p and
load_killed_in_block_p. Nonzero if mems_conflict_for_gcse_p finds a
conflict between two memory references. */
static int gcse_mems_conflict_p;
/* Used for communication between mems_conflict_for_gcse_p and
load_killed_in_block_p. A memory reference for a load instruction,
mems_conflict_for_gcse_p will see if a memory store conflicts with
this memory load. */
static rtx gcse_mem_operand;
/* DEST is the output of an instruction. If it is a memory reference, and
possibly conflicts with the load found in gcse_mem_operand, then set
gcse_mems_conflict_p to a nonzero value. */
static void
mems_conflict_for_gcse_p (rtx dest, rtx setter ATTRIBUTE_UNUSED,
void *data ATTRIBUTE_UNUSED)
{
while (GET_CODE (dest) == SUBREG
|| GET_CODE (dest) == ZERO_EXTRACT
|| GET_CODE (dest) == SIGN_EXTRACT
|| GET_CODE (dest) == STRICT_LOW_PART)
dest = XEXP (dest, 0);
/* If DEST is not a MEM, then it will not conflict with the load. Note
that function calls are assumed to clobber memory, but are handled
elsewhere. */
if (GET_CODE (dest) != MEM)
return;
/* If we are setting a MEM in our list of specially recognized MEMs,
don't mark as killed this time. */
if (expr_equiv_p (dest, gcse_mem_operand) && pre_ldst_mems != NULL)
{
if (!find_rtx_in_ldst (dest))
gcse_mems_conflict_p = 1;
return;
}
if (true_dependence (dest, GET_MODE (dest), gcse_mem_operand,
rtx_addr_varies_p))
gcse_mems_conflict_p = 1;
}
/* Return nonzero if the expression in X (a memory reference) is killed
in block BB before or after the insn with the CUID in UID_LIMIT.
AVAIL_P is nonzero for kills after UID_LIMIT, and zero for kills
before UID_LIMIT.
To check the entire block, set UID_LIMIT to max_uid + 1 and
AVAIL_P to 0. */
static int
load_killed_in_block_p (basic_block bb, int uid_limit, rtx x, int avail_p)
{
rtx list_entry = modify_mem_list[bb->index];
while (list_entry)
{
rtx setter;
/* Ignore entries in the list that do not apply. */
if ((avail_p
&& INSN_CUID (XEXP (list_entry, 0)) < uid_limit)
|| (! avail_p
&& INSN_CUID (XEXP (list_entry, 0)) > uid_limit))
{
list_entry = XEXP (list_entry, 1);
continue;
}
setter = XEXP (list_entry, 0);
/* If SETTER is a call everything is clobbered. Note that calls
to pure functions are never put on the list, so we need not
worry about them. */
if (GET_CODE (setter) == CALL_INSN)
return 1;
/* SETTER must be an INSN of some kind that sets memory. Call
note_stores to examine each hunk of memory that is modified.
The note_stores interface is pretty limited, so we have to
communicate via global variables. Yuk. */
gcse_mem_operand = x;
gcse_mems_conflict_p = 0;
note_stores (PATTERN (setter), mems_conflict_for_gcse_p, NULL);
if (gcse_mems_conflict_p)
return 1;
list_entry = XEXP (list_entry, 1);
}
return 0;
}
/* Return nonzero if the operands of expression X are unchanged from
the start of INSN's basic block up to but not including INSN. */
static int
oprs_anticipatable_p (rtx x, rtx insn)
{
return oprs_unchanged_p (x, insn, 0);
}
/* Return nonzero if the operands of expression X are unchanged from
INSN to the end of INSN's basic block. */
static int
oprs_available_p (rtx x, rtx insn)
{
return oprs_unchanged_p (x, insn, 1);
}
/* Hash expression X.
MODE is only used if X is a CONST_INT. DO_NOT_RECORD_P is a boolean
indicating if a volatile operand is found or if the expression contains
something we don't want to insert in the table. HASH_TABLE_SIZE is
the current size of the hash table to be probed.
??? One might want to merge this with canon_hash. Later. */
static unsigned int
hash_expr (rtx x, enum machine_mode mode, int *do_not_record_p,
int hash_table_size)
{
unsigned int hash;
*do_not_record_p = 0;
hash = hash_expr_1 (x, mode, do_not_record_p);
return hash % hash_table_size;
}
/* Hash a string. Just add its bytes up. */
static inline unsigned
hash_string_1 (const char *ps)
{
unsigned hash = 0;
const unsigned char *p = (const unsigned char *) ps;
if (p)
while (*p)
hash += *p++;
return hash;
}
/* Subroutine of hash_expr to do the actual work. */
static unsigned int
hash_expr_1 (rtx x, enum machine_mode mode, int *do_not_record_p)
{
int i, j;
unsigned hash = 0;
enum rtx_code code;
const char *fmt;
/* Used to turn recursion into iteration. We can't rely on GCC's
tail-recursion elimination since we need to keep accumulating values
in HASH. */
if (x == 0)
return hash;
repeat:
code = GET_CODE (x);
switch (code)
{
case REG:
hash += ((unsigned int) REG << 7) + REGNO (x);
return hash;
case CONST_INT:
hash += (((unsigned int) CONST_INT << 7) + (unsigned int) mode
+ (unsigned int) INTVAL (x));
return hash;
case CONST_DOUBLE:
/* This is like the general case, except that it only counts
the integers representing the constant. */
hash += (unsigned int) code + (unsigned int) GET_MODE (x);
if (GET_MODE (x) != VOIDmode)
for (i = 2; i < GET_RTX_LENGTH (CONST_DOUBLE); i++)
hash += (unsigned int) XWINT (x, i);
else
hash += ((unsigned int) CONST_DOUBLE_LOW (x)
+ (unsigned int) CONST_DOUBLE_HIGH (x));
return hash;
case CONST_VECTOR:
{
int units;
rtx elt;
units = CONST_VECTOR_NUNITS (x);
for (i = 0; i < units; ++i)
{
elt = CONST_VECTOR_ELT (x, i);
hash += hash_expr_1 (elt, GET_MODE (elt), do_not_record_p);
}
return hash;
}
/* Assume there is only one rtx object for any given label. */
case LABEL_REF:
/* We don't hash on the address of the CODE_LABEL to avoid bootstrap
differences and differences between each stage's debugging dumps. */
hash += (((unsigned int) LABEL_REF << 7)
+ CODE_LABEL_NUMBER (XEXP (x, 0)));
return hash;
case SYMBOL_REF:
{
/* Don't hash on the symbol's address to avoid bootstrap differences.
Different hash values may cause expressions to be recorded in
different orders and thus different registers to be used in the
final assembler. This also avoids differences in the dump files
between various stages. */
unsigned int h = 0;
const unsigned char *p = (const unsigned char *) XSTR (x, 0);
while (*p)
h += (h << 7) + *p++; /* ??? revisit */
hash += ((unsigned int) SYMBOL_REF << 7) + h;
return hash;
}
case MEM:
if (MEM_VOLATILE_P (x))
{
*do_not_record_p = 1;
return 0;
}
hash += (unsigned int) MEM;
/* We used alias set for hashing, but this is not good, since the alias
set may differ in -fprofile-arcs and -fbranch-probabilities compilation
causing the profiles to fail to match. */
x = XEXP (x, 0);
goto repeat;
case PRE_DEC:
case PRE_INC:
case POST_DEC:
case POST_INC:
case PC:
case CC0:
case CALL:
case UNSPEC_VOLATILE:
*do_not_record_p = 1;
return 0;
case ASM_OPERANDS:
if (MEM_VOLATILE_P (x))
{
*do_not_record_p = 1;
return 0;
}
else
{
/* We don't want to take the filename and line into account. */
hash += (unsigned) code + (unsigned) GET_MODE (x)
+ hash_string_1 (ASM_OPERANDS_TEMPLATE (x))
+ hash_string_1 (ASM_OPERANDS_OUTPUT_CONSTRAINT (x))
+ (unsigned) ASM_OPERANDS_OUTPUT_IDX (x);
if (ASM_OPERANDS_INPUT_LENGTH (x))
{
for (i = 1; i < ASM_OPERANDS_INPUT_LENGTH (x); i++)
{
hash += (hash_expr_1 (ASM_OPERANDS_INPUT (x, i),
GET_MODE (ASM_OPERANDS_INPUT (x, i)),
do_not_record_p)
+ hash_string_1 (ASM_OPERANDS_INPUT_CONSTRAINT
(x, i)));
}
hash += hash_string_1 (ASM_OPERANDS_INPUT_CONSTRAINT (x, 0));
x = ASM_OPERANDS_INPUT (x, 0);
mode = GET_MODE (x);
goto repeat;
}
return hash;
}
default:
break;
}
hash += (unsigned) code + (unsigned) GET_MODE (x);
for (i = GET_RTX_LENGTH (code) - 1, fmt = GET_RTX_FORMAT (code); i >= 0; i--)
{
if (fmt[i] == 'e')
{
/* If we are about to do the last recursive call
needed at this level, change it into iteration.
This function is called enough to be worth it. */
if (i == 0)
{
x = XEXP (x, i);
goto repeat;
}
hash += hash_expr_1 (XEXP (x, i), 0, do_not_record_p);
if (*do_not_record_p)
return 0;
}
else if (fmt[i] == 'E')
for (j = 0; j < XVECLEN (x, i); j++)
{
hash += hash_expr_1 (XVECEXP (x, i, j), 0, do_not_record_p);
if (*do_not_record_p)
return 0;
}
else if (fmt[i] == 's')
hash += hash_string_1 (XSTR (x, i));
else if (fmt[i] == 'i')
hash += (unsigned int) XINT (x, i);
else
abort ();
}
return hash;
}
/* Hash a set of register REGNO.
Sets are hashed on the register that is set. This simplifies the PRE copy
propagation code.
??? May need to make things more elaborate. Later, as necessary. */
static unsigned int
hash_set (int regno, int hash_table_size)
{
unsigned int hash;
hash = regno;
return hash % hash_table_size;
}
/* Return nonzero if exp1 is equivalent to exp2.
??? Borrowed from cse.c. Might want to remerge with cse.c. Later. */
static int
expr_equiv_p (rtx x, rtx y)
{
int i, j;
enum rtx_code code;
const char *fmt;
if (x == y)
return 1;
if (x == 0 || y == 0)
return 0;
code = GET_CODE (x);
if (code != GET_CODE (y))
return 0;
/* (MULT:SI x y) and (MULT:HI x y) are NOT equivalent. */
if (GET_MODE (x) != GET_MODE (y))
return 0;
switch (code)
{
case PC:
case CC0:
case CONST_INT:
return 0;
case LABEL_REF:
return XEXP (x, 0) == XEXP (y, 0);
case SYMBOL_REF:
return XSTR (x, 0) == XSTR (y, 0);
case REG:
return REGNO (x) == REGNO (y);
case MEM:
/* Can't merge two expressions in different alias sets, since we can
decide that the expression is transparent in a block when it isn't,
due to it being set with the different alias set. */
if (MEM_ALIAS_SET (x) != MEM_ALIAS_SET (y))
return 0;
/* A volatile mem should not be considered equivalent to any other. */
if (MEM_VOLATILE_P (x) || MEM_VOLATILE_P (y))
return 0;
break;
/* For commutative operations, check both orders. */
case PLUS:
case MULT:
case AND:
case IOR:
case XOR:
case NE:
case EQ:
return ((expr_equiv_p (XEXP (x, 0), XEXP (y, 0))
&& expr_equiv_p (XEXP (x, 1), XEXP (y, 1)))
|| (expr_equiv_p (XEXP (x, 0), XEXP (y, 1))
&& expr_equiv_p (XEXP (x, 1), XEXP (y, 0))));
case ASM_OPERANDS:
/* We don't use the generic code below because we want to
disregard filename and line numbers. */
/* A volatile asm isn't equivalent to any other. */
if (MEM_VOLATILE_P (x) || MEM_VOLATILE_P (y))
return 0;
if (GET_MODE (x) != GET_MODE (y)
|| strcmp (ASM_OPERANDS_TEMPLATE (x), ASM_OPERANDS_TEMPLATE (y))
|| strcmp (ASM_OPERANDS_OUTPUT_CONSTRAINT (x),
ASM_OPERANDS_OUTPUT_CONSTRAINT (y))
|| ASM_OPERANDS_OUTPUT_IDX (x) != ASM_OPERANDS_OUTPUT_IDX (y)
|| ASM_OPERANDS_INPUT_LENGTH (x) != ASM_OPERANDS_INPUT_LENGTH (y))
return 0;
if (ASM_OPERANDS_INPUT_LENGTH (x))
{
for (i = ASM_OPERANDS_INPUT_LENGTH (x) - 1; i >= 0; i--)
if (! expr_equiv_p (ASM_OPERANDS_INPUT (x, i),
ASM_OPERANDS_INPUT (y, i))
|| strcmp (ASM_OPERANDS_INPUT_CONSTRAINT (x, i),
ASM_OPERANDS_INPUT_CONSTRAINT (y, i)))
return 0;
}
return 1;
default:
break;
}
/* Compare the elements. If any pair of corresponding elements
fail to match, return 0 for the whole thing. */
fmt = GET_RTX_FORMAT (code);
for (i = GET_RTX_LENGTH (code) - 1; i >= 0; i--)
{
switch (fmt[i])
{
case 'e':
if (! expr_equiv_p (XEXP (x, i), XEXP (y, i)))
return 0;
break;
case 'E':
if (XVECLEN (x, i) != XVECLEN (y, i))
return 0;
for (j = 0; j < XVECLEN (x, i); j++)
if (! expr_equiv_p (XVECEXP (x, i, j), XVECEXP (y, i, j)))
return 0;
break;
case 's':
if (strcmp (XSTR (x, i), XSTR (y, i)))
return 0;
break;
case 'i':
if (XINT (x, i) != XINT (y, i))
return 0;
break;
case 'w':
if (XWINT (x, i) != XWINT (y, i))
return 0;
break;
case '0':
break;
default:
abort ();
}
}
return 1;
}
/* Insert expression X in INSN in the hash TABLE.
If it is already present, record it as the last occurrence in INSN's
basic block.
MODE is the mode of the value X is being stored into.
It is only used if X is a CONST_INT.
ANTIC_P is nonzero if X is an anticipatable expression.
AVAIL_P is nonzero if X is an available expression. */
static void
insert_expr_in_table (rtx x, enum machine_mode mode, rtx insn, int antic_p,
int avail_p, struct hash_table *table)
{
int found, do_not_record_p;
unsigned int hash;
struct expr *cur_expr, *last_expr = NULL;
struct occr *antic_occr, *avail_occr;
struct occr *last_occr = NULL;
hash = hash_expr (x, mode, &do_not_record_p, table->size);
/* Do not insert expression in table if it contains volatile operands,
or if hash_expr determines the expression is something we don't want
to or can't handle. */
if (do_not_record_p)
return;
cur_expr = table->table[hash];
found = 0;
while (cur_expr && 0 == (found = expr_equiv_p (cur_expr->expr, x)))
{
/* If the expression isn't found, save a pointer to the end of
the list. */
last_expr = cur_expr;
cur_expr = cur_expr->next_same_hash;
}
if (! found)
{
cur_expr = gcse_alloc (sizeof (struct expr));
bytes_used += sizeof (struct expr);
if (table->table[hash] == NULL)
/* This is the first pattern that hashed to this index. */
table->table[hash] = cur_expr;
else
/* Add EXPR to end of this hash chain. */
last_expr->next_same_hash = cur_expr;
/* Set the fields of the expr element. */
cur_expr->expr = x;
cur_expr->bitmap_index = table->n_elems++;
cur_expr->next_same_hash = NULL;
cur_expr->antic_occr = NULL;
cur_expr->avail_occr = NULL;
}
/* Now record the occurrence(s). */
if (antic_p)
{
antic_occr = cur_expr->antic_occr;
/* Search for another occurrence in the same basic block. */
while (antic_occr && BLOCK_NUM (antic_occr->insn) != BLOCK_NUM (insn))
{
/* If an occurrence isn't found, save a pointer to the end of
the list. */
last_occr = antic_occr;
antic_occr = antic_occr->next;
}
if (antic_occr)
/* Found another instance of the expression in the same basic block.
Prefer the currently recorded one. We want the first one in the
block and the block is scanned from start to end. */
; /* nothing to do */
else
{
/* First occurrence of this expression in this basic block. */
antic_occr = gcse_alloc (sizeof (struct occr));
bytes_used += sizeof (struct occr);
/* First occurrence of this expression in any block? */
if (cur_expr->antic_occr == NULL)
cur_expr->antic_occr = antic_occr;
else
last_occr->next = antic_occr;
antic_occr->insn = insn;
antic_occr->next = NULL;
}
}
if (avail_p)
{
avail_occr = cur_expr->avail_occr;
/* Search for another occurrence in the same basic block. */
while (avail_occr && BLOCK_NUM (avail_occr->insn) != BLOCK_NUM (insn))
{
/* If an occurrence isn't found, save a pointer to the end of
the list. */
last_occr = avail_occr;
avail_occr = avail_occr->next;
}
if (avail_occr)
/* Found another instance of the expression in the same basic block.
Prefer this occurrence to the currently recorded one. We want
the last one in the block and the block is scanned from start
to end. */
avail_occr->insn = insn;
else
{
/* First occurrence of this expression in this basic block. */
avail_occr = gcse_alloc (sizeof (struct occr));
bytes_used += sizeof (struct occr);
/* First occurrence of this expression in any block? */
if (cur_expr->avail_occr == NULL)
cur_expr->avail_occr = avail_occr;
else
last_occr->next = avail_occr;
avail_occr->insn = insn;
avail_occr->next = NULL;
}
}
}
/* Insert pattern X in INSN in the hash table.
X is a SET of a reg to either another reg or a constant.
If it is already present, record it as the last occurrence in INSN's
basic block. */
static void
insert_set_in_table (rtx x, rtx insn, struct hash_table *table)
{
int found;
unsigned int hash;
struct expr *cur_expr, *last_expr = NULL;
struct occr *cur_occr, *last_occr = NULL;
if (GET_CODE (x) != SET
|| GET_CODE (SET_DEST (x)) != REG)
abort ();
hash = hash_set (REGNO (SET_DEST (x)), table->size);
cur_expr = table->table[hash];
found = 0;
while (cur_expr && 0 == (found = expr_equiv_p (cur_expr->expr, x)))
{
/* If the expression isn't found, save a pointer to the end of
the list. */
last_expr = cur_expr;
cur_expr = cur_expr->next_same_hash;
}
if (! found)
{
cur_expr = gcse_alloc (sizeof (struct expr));
bytes_used += sizeof (struct expr);
if (table->table[hash] == NULL)
/* This is the first pattern that hashed to this index. */
table->table[hash] = cur_expr;
else
/* Add EXPR to end of this hash chain. */
last_expr->next_same_hash = cur_expr;
/* Set the fields of the expr element.
We must copy X because it can be modified when copy propagation is
performed on its operands. */
cur_expr->expr = copy_rtx (x);
cur_expr->bitmap_index = table->n_elems++;
cur_expr->next_same_hash = NULL;
cur_expr->antic_occr = NULL;
cur_expr->avail_occr = NULL;
}
/* Now record the occurrence. */
cur_occr = cur_expr->avail_occr;
/* Search for another occurrence in the same basic block. */
while (cur_occr && BLOCK_NUM (cur_occr->insn) != BLOCK_NUM (insn))
{
/* If an occurrence isn't found, save a pointer to the end of
the list. */
last_occr = cur_occr;
cur_occr = cur_occr->next;
}
if (cur_occr)
/* Found another instance of the expression in the same basic block.
Prefer this occurrence to the currently recorded one. We want the
last one in the block and the block is scanned from start to end. */
cur_occr->insn = insn;
else
{
/* First occurrence of this expression in this basic block. */
cur_occr = gcse_alloc (sizeof (struct occr));
bytes_used += sizeof (struct occr);
/* First occurrence of this expression in any block? */
if (cur_expr->avail_occr == NULL)
cur_expr->avail_occr = cur_occr;
else
last_occr->next = cur_occr;
cur_occr->insn = insn;
cur_occr->next = NULL;
}
}
/* Determine whether the rtx X should be treated as a constant for
the purposes of GCSE's constant propagation. */
static bool
gcse_constant_p (rtx x)
{
/* Consider a COMPARE of two integers constant. */
if (GET_CODE (x) == COMPARE
&& GET_CODE (XEXP (x, 0)) == CONST_INT
&& GET_CODE (XEXP (x, 1)) == CONST_INT)
return true;
/* Consider a COMPARE of the same registers is a constant
if they are not floating point registers. */
if (GET_CODE(x) == COMPARE
&& GET_CODE (XEXP (x, 0)) == REG
&& GET_CODE (XEXP (x, 1)) == REG
&& REGNO (XEXP (x, 0)) == REGNO (XEXP (x, 1))
&& ! FLOAT_MODE_P (GET_MODE (XEXP (x, 0)))
&& ! FLOAT_MODE_P (GET_MODE (XEXP (x, 1))))
return true;
if (GET_CODE (x) == CONSTANT_P_RTX)
return false;
return CONSTANT_P (x);
}
/* Scan pattern PAT of INSN and add an entry to the hash TABLE (set or
expression one). */
static void
hash_scan_set (rtx pat, rtx insn, struct hash_table *table)
{
rtx src = SET_SRC (pat);
rtx dest = SET_DEST (pat);
rtx note;
if (GET_CODE (src) == CALL)
hash_scan_call (src, insn, table);
else if (GET_CODE (dest) == REG)
{
unsigned int regno = REGNO (dest);
rtx tmp;
/* If this is a single set and we are doing constant propagation,
see if a REG_NOTE shows this equivalent to a constant. */
if (table->set_p && (note = find_reg_equal_equiv_note (insn)) != 0
&& gcse_constant_p (XEXP (note, 0)))
src = XEXP (note, 0), pat = gen_rtx_SET (VOIDmode, dest, src);
/* Only record sets of pseudo-regs in the hash table. */
if (! table->set_p
&& regno >= FIRST_PSEUDO_REGISTER
/* Don't GCSE something if we can't do a reg/reg copy. */
&& can_copy_p (GET_MODE (dest))
/* GCSE commonly inserts instruction after the insn. We can't
do that easily for EH_REGION notes so disable GCSE on these
for now. */
&& !find_reg_note (insn, REG_EH_REGION, NULL_RTX)
/* Is SET_SRC something we want to gcse? */
&& want_to_gcse_p (src)
/* Don't CSE a nop. */
&& ! set_noop_p (pat)
/* Don't GCSE if it has attached REG_EQUIV note.
At this point this only function parameters should have
REG_EQUIV notes and if the argument slot is used somewhere
explicitly, it means address of parameter has been taken,
so we should not extend the lifetime of the pseudo. */
&& ((note = find_reg_note (insn, REG_EQUIV, NULL_RTX)) == 0
|| GET_CODE (XEXP (note, 0)) != MEM))
{
/* An expression is not anticipatable if its operands are
modified before this insn or if this is not the only SET in
this insn. */
int antic_p = oprs_anticipatable_p (src, insn) && single_set (insn);
/* An expression is not available if its operands are
subsequently modified, including this insn. It's also not
available if this is a branch, because we can't insert
a set after the branch. */
int avail_p = (oprs_available_p (src, insn)
&& ! JUMP_P (insn));
insert_expr_in_table (src, GET_MODE (dest), insn, antic_p, avail_p, table);
}
/* Record sets for constant/copy propagation. */
else if (table->set_p
&& regno >= FIRST_PSEUDO_REGISTER
&& ((GET_CODE (src) == REG
&& REGNO (src) >= FIRST_PSEUDO_REGISTER
&& can_copy_p (GET_MODE (dest))
&& REGNO (src) != regno)
|| gcse_constant_p (src))
/* A copy is not available if its src or dest is subsequently
modified. Here we want to search from INSN+1 on, but
oprs_available_p searches from INSN on. */
&& (insn == BB_END (BLOCK_FOR_INSN (insn))
|| ((tmp = next_nonnote_insn (insn)) != NULL_RTX
&& oprs_available_p (pat, tmp))))
insert_set_in_table (pat, insn, table);
}
/* In case of store we want to consider the memory value as available in
the REG stored in that memory. This makes it possible to remove
redundant loads from due to stores to the same location. */
else if (flag_gcse_las && GET_CODE (src) == REG && GET_CODE (dest) == MEM)
{
unsigned int regno = REGNO (src);
/* Do not do this for constant/copy propagation. */
if (! table->set_p
/* Only record sets of pseudo-regs in the hash table. */
&& regno >= FIRST_PSEUDO_REGISTER
/* Don't GCSE something if we can't do a reg/reg copy. */
&& can_copy_p (GET_MODE (src))
/* GCSE commonly inserts instruction after the insn. We can't
do that easily for EH_REGION notes so disable GCSE on these
for now. */
&& ! find_reg_note (insn, REG_EH_REGION, NULL_RTX)
/* Is SET_DEST something we want to gcse? */
&& want_to_gcse_p (dest)
/* Don't CSE a nop. */
&& ! set_noop_p (pat)
/* Don't GCSE if it has attached REG_EQUIV note.
At this point this only function parameters should have
REG_EQUIV notes and if the argument slot is used somewhere
explicitly, it means address of parameter has been taken,
so we should not extend the lifetime of the pseudo. */
&& ((note = find_reg_note (insn, REG_EQUIV, NULL_RTX)) == 0
|| GET_CODE (XEXP (note, 0)) != MEM))
{
/* Stores are never anticipatable. */
int antic_p = 0;
/* An expression is not available if its operands are
subsequently modified, including this insn. It's also not
available if this is a branch, because we can't insert
a set after the branch. */
int avail_p = oprs_available_p (dest, insn)
&& ! JUMP_P (insn);
/* Record the memory expression (DEST) in the hash table. */
insert_expr_in_table (dest, GET_MODE (dest), insn,
antic_p, avail_p, table);
}
}
}
static void
hash_scan_clobber (rtx x ATTRIBUTE_UNUSED, rtx insn ATTRIBUTE_UNUSED,
struct hash_table *table ATTRIBUTE_UNUSED)
{
/* Currently nothing to do. */
}
static void
hash_scan_call (rtx x ATTRIBUTE_UNUSED, rtx insn ATTRIBUTE_UNUSED,
struct hash_table *table ATTRIBUTE_UNUSED)
{
/* Currently nothing to do. */
}
/* Process INSN and add hash table entries as appropriate.
Only available expressions that set a single pseudo-reg are recorded.
Single sets in a PARALLEL could be handled, but it's an extra complication
that isn't dealt with right now. The trick is handling the CLOBBERs that
are also in the PARALLEL. Later.
If SET_P is nonzero, this is for the assignment hash table,
otherwise it is for the expression hash table.
If IN_LIBCALL_BLOCK nonzero, we are in a libcall block, and should
not record any expressions. */
static void
hash_scan_insn (rtx insn, struct hash_table *table, int in_libcall_block)
{
rtx pat = PATTERN (insn);
int i;
if (in_libcall_block)
return;
/* Pick out the sets of INSN and for other forms of instructions record
what's been modified. */
if (GET_CODE (pat) == SET)
hash_scan_set (pat, insn, table);
else if (GET_CODE (pat) == PARALLEL)
for (i = 0; i < XVECLEN (pat, 0); i++)
{
rtx x = XVECEXP (pat, 0, i);
if (GET_CODE (x) == SET)
hash_scan_set (x, insn, table);
else if (GET_CODE (x) == CLOBBER)
hash_scan_clobber (x, insn, table);
else if (GET_CODE (x) == CALL)
hash_scan_call (x, insn, table);
}
else if (GET_CODE (pat) == CLOBBER)
hash_scan_clobber (pat, insn, table);
else if (GET_CODE (pat) == CALL)
hash_scan_call (pat, insn, table);
}
static void
dump_hash_table (FILE *file, const char *name, struct hash_table *table)
{
int i;
/* Flattened out table, so it's printed in proper order. */
struct expr **flat_table;
unsigned int *hash_val;
struct expr *expr;
flat_table = xcalloc (table->n_elems, sizeof (struct expr *));
hash_val = xmalloc (table->n_elems * sizeof (unsigned int));
for (i = 0; i < (int) table->size; i++)
for (expr = table->table[i]; expr != NULL; expr = expr->next_same_hash)
{
flat_table[expr->bitmap_index] = expr;
hash_val[expr->bitmap_index] = i;
}
fprintf (file, "%s hash table (%d buckets, %d entries)\n",
name, table->size, table->n_elems);
for (i = 0; i < (int) table->n_elems; i++)
if (flat_table[i] != 0)
{
expr = flat_table[i];
fprintf (file, "Index %d (hash value %d)\n ",
expr->bitmap_index, hash_val[i]);
print_rtl (file, expr->expr);
fprintf (file, "\n");
}
fprintf (file, "\n");
free (flat_table);
free (hash_val);
}
/* Record register first/last/block set information for REGNO in INSN.
first_set records the first place in the block where the register
is set and is used to compute "anticipatability".
last_set records the last place in the block where the register
is set and is used to compute "availability".
last_bb records the block for which first_set and last_set are
valid, as a quick test to invalidate them.
reg_set_in_block records whether the register is set in the block
and is used to compute "transparency". */
static void
record_last_reg_set_info (rtx insn, int regno)
{
struct reg_avail_info *info = &reg_avail_info[regno];
int cuid = INSN_CUID (insn);
info->last_set = cuid;
if (info->last_bb != current_bb)
{
info->last_bb = current_bb;
info->first_set = cuid;
SET_BIT (reg_set_in_block[current_bb->index], regno);
}
}
/* Record all of the canonicalized MEMs of record_last_mem_set_info's insn.
Note we store a pair of elements in the list, so they have to be
taken off pairwise. */
static void
canon_list_insert (rtx dest ATTRIBUTE_UNUSED, rtx unused1 ATTRIBUTE_UNUSED,
void * v_insn)
{
rtx dest_addr, insn;
int bb;
while (GET_CODE (dest) == SUBREG
|| GET_CODE (dest) == ZERO_EXTRACT
|| GET_CODE (dest) == SIGN_EXTRACT
|| GET_CODE (dest) == STRICT_LOW_PART)
dest = XEXP (dest, 0);
/* If DEST is not a MEM, then it will not conflict with a load. Note
that function calls are assumed to clobber memory, but are handled
elsewhere. */
if (GET_CODE (dest) != MEM)
return;
dest_addr = get_addr (XEXP (dest, 0));
dest_addr = canon_rtx (dest_addr);
insn = (rtx) v_insn;
bb = BLOCK_NUM (insn);
canon_modify_mem_list[bb] =
alloc_EXPR_LIST (VOIDmode, dest_addr, canon_modify_mem_list[bb]);
canon_modify_mem_list[bb] =
alloc_EXPR_LIST (VOIDmode, dest, canon_modify_mem_list[bb]);
bitmap_set_bit (canon_modify_mem_list_set, bb);
}
/* Record memory modification information for INSN. We do not actually care
about the memory location(s) that are set, or even how they are set (consider
a CALL_INSN). We merely need to record which insns modify memory. */
static void
record_last_mem_set_info (rtx insn)
{
int bb = BLOCK_NUM (insn);
/* load_killed_in_block_p will handle the case of calls clobbering
everything. */
modify_mem_list[bb] = alloc_INSN_LIST (insn, modify_mem_list[bb]);
bitmap_set_bit (modify_mem_list_set, bb);
if (GET_CODE (insn) == CALL_INSN)
{
/* Note that traversals of this loop (other than for free-ing)
will break after encountering a CALL_INSN. So, there's no
need to insert a pair of items, as canon_list_insert does. */
canon_modify_mem_list[bb] =
alloc_INSN_LIST (insn, canon_modify_mem_list[bb]);
bitmap_set_bit (canon_modify_mem_list_set, bb);
}
else
note_stores (PATTERN (insn), canon_list_insert, (void*) insn);
}
/* Called from compute_hash_table via note_stores to handle one
SET or CLOBBER in an insn. DATA is really the instruction in which
the SET is taking place. */
static void
record_last_set_info (rtx dest, rtx setter ATTRIBUTE_UNUSED, void *data)
{
rtx last_set_insn = (rtx) data;
if (GET_CODE (dest) == SUBREG)
dest = SUBREG_REG (dest);
if (GET_CODE (dest) == REG)
record_last_reg_set_info (last_set_insn, REGNO (dest));
else if (GET_CODE (dest) == MEM
/* Ignore pushes, they clobber nothing. */
&& ! push_operand (dest, GET_MODE (dest)))
record_last_mem_set_info (last_set_insn);
}
/* Top level function to create an expression or assignment hash table.
Expression entries are placed in the hash table if
- they are of the form (set (pseudo-reg) src),
- src is something we want to perform GCSE on,
- none of the operands are subsequently modified in the block
Assignment entries are placed in the hash table if
- they are of the form (set (pseudo-reg) src),
- src is something we want to perform const/copy propagation on,
- none of the operands or target are subsequently modified in the block
Currently src must be a pseudo-reg or a const_int.
TABLE is the table computed. */
static void
compute_hash_table_work (struct hash_table *table)
{
unsigned int i;
/* While we compute the hash table we also compute a bit array of which
registers are set in which blocks.
??? This isn't needed during const/copy propagation, but it's cheap to
compute. Later. */
sbitmap_vector_zero (reg_set_in_block, last_basic_block);
/* re-Cache any INSN_LIST nodes we have allocated. */
clear_modify_mem_tables ();
/* Some working arrays used to track first and last set in each block. */
reg_avail_info = gmalloc (max_gcse_regno * sizeof (struct reg_avail_info));
for (i = 0; i < max_gcse_regno; ++i)
reg_avail_info[i].last_bb = NULL;
FOR_EACH_BB (current_bb)
{
rtx insn;
unsigned int regno;
int in_libcall_block;
/* First pass over the instructions records information used to
determine when registers and memory are first and last set.
??? hard-reg reg_set_in_block computation
could be moved to compute_sets since they currently don't change. */
for (insn = BB_HEAD (current_bb);
insn && insn != NEXT_INSN (BB_END (current_bb));
insn = NEXT_INSN (insn))
{
if (! INSN_P (insn))
continue;
if (GET_CODE (insn) == CALL_INSN)
{
bool clobbers_all = false;
#ifdef NON_SAVING_SETJMP
if (NON_SAVING_SETJMP
&& find_reg_note (insn, REG_SETJMP, NULL_RTX))
clobbers_all = true;
#endif
for (regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
if (clobbers_all
|| TEST_HARD_REG_BIT (regs_invalidated_by_call, regno))
record_last_reg_set_info (insn, regno);
mark_call (insn);
}
note_stores (PATTERN (insn), record_last_set_info, insn);
}
/* Insert implicit sets in the hash table. */
if (table->set_p
&& implicit_sets[current_bb->index] != NULL_RTX)
hash_scan_set (implicit_sets[current_bb->index],
BB_HEAD (current_bb), table);
/* The next pass builds the hash table. */
for (insn = BB_HEAD (current_bb), in_libcall_block = 0;
insn && insn != NEXT_INSN (BB_END (current_bb));
insn = NEXT_INSN (insn))
if (INSN_P (insn))
{
if (find_reg_note (insn, REG_LIBCALL, NULL_RTX))
in_libcall_block = 1;
else if (table->set_p && find_reg_note (insn, REG_RETVAL, NULL_RTX))
in_libcall_block = 0;
hash_scan_insn (insn, table, in_libcall_block);
if (!table->set_p && find_reg_note (insn, REG_RETVAL, NULL_RTX))
in_libcall_block = 0;
}
}
free (reg_avail_info);
reg_avail_info = NULL;
}
/* Allocate space for the set/expr hash TABLE.
N_INSNS is the number of instructions in the function.
It is used to determine the number of buckets to use.
SET_P determines whether set or expression table will
be created. */
static void
alloc_hash_table (int n_insns, struct hash_table *table, int set_p)
{
int n;
table->size = n_insns / 4;
if (table->size < 11)
table->size = 11;
/* Attempt to maintain efficient use of hash table.
Making it an odd number is simplest for now.
??? Later take some measurements. */
table->size |= 1;
n = table->size * sizeof (struct expr *);
table->table = gmalloc (n);
table->set_p = set_p;
}
/* Free things allocated by alloc_hash_table. */
static void
free_hash_table (struct hash_table *table)
{
free (table->table);
}
/* Compute the hash TABLE for doing copy/const propagation or
expression hash table. */
static void
compute_hash_table (struct hash_table *table)
{
/* Initialize count of number of entries in hash table. */
table->n_elems = 0;
memset (table->table, 0, table->size * sizeof (struct expr *));
compute_hash_table_work (table);
}
/* Expression tracking support. */
/* Lookup pattern PAT in the expression TABLE.
The result is a pointer to the table entry, or NULL if not found. */
static struct expr *
lookup_expr (rtx pat, struct hash_table *table)
{
int do_not_record_p;
unsigned int hash = hash_expr (pat, GET_MODE (pat), &do_not_record_p,
table->size);
struct expr *expr;
if (do_not_record_p)
return NULL;
expr = table->table[hash];
while (expr && ! expr_equiv_p (expr->expr, pat))
expr = expr->next_same_hash;
return expr;
}
/* Lookup REGNO in the set TABLE. The result is a pointer to the
table entry, or NULL if not found. */
static struct expr *
lookup_set (unsigned int regno, struct hash_table *table)
{
unsigned int hash = hash_set (regno, table->size);
struct expr *expr;
expr = table->table[hash];
while (expr && REGNO (SET_DEST (expr->expr)) != regno)
expr = expr->next_same_hash;
return expr;
}
/* Return the next entry for REGNO in list EXPR. */
static struct expr *
next_set (unsigned int regno, struct expr *expr)
{
do
expr = expr->next_same_hash;
while (expr && REGNO (SET_DEST (expr->expr)) != regno);
return expr;
}
/* Like free_INSN_LIST_list or free_EXPR_LIST_list, except that the node
types may be mixed. */
static void
free_insn_expr_list_list (rtx *listp)
{
rtx list, next;
for (list = *listp; list ; list = next)
{
next = XEXP (list, 1);
if (GET_CODE (list) == EXPR_LIST)
free_EXPR_LIST_node (list);
else
free_INSN_LIST_node (list);
}
*listp = NULL;
}
/* Clear canon_modify_mem_list and modify_mem_list tables. */
static void
clear_modify_mem_tables (void)
{
int i;
EXECUTE_IF_SET_IN_BITMAP
(modify_mem_list_set, 0, i, free_INSN_LIST_list (modify_mem_list + i));
bitmap_clear (modify_mem_list_set);
EXECUTE_IF_SET_IN_BITMAP
(canon_modify_mem_list_set, 0, i,
free_insn_expr_list_list (canon_modify_mem_list + i));
bitmap_clear (canon_modify_mem_list_set);
}
/* Release memory used by modify_mem_list_set and canon_modify_mem_list_set. */
static void
free_modify_mem_tables (void)
{
clear_modify_mem_tables ();
free (modify_mem_list);
free (canon_modify_mem_list);
modify_mem_list = 0;
canon_modify_mem_list = 0;
}
/* Reset tables used to keep track of what's still available [since the
start of the block]. */
static void
reset_opr_set_tables (void)
{
/* Maintain a bitmap of which regs have been set since beginning of
the block. */
CLEAR_REG_SET (reg_set_bitmap);
/* Also keep a record of the last instruction to modify memory.
For now this is very trivial, we only record whether any memory
location has been modified. */
clear_modify_mem_tables ();
}
/* Return nonzero if the operands of X are not set before INSN in
INSN's basic block. */
static int
oprs_not_set_p (rtx x, rtx insn)
{
int i, j;
enum rtx_code code;
const char *fmt;
if (x == 0)
return 1;
code = GET_CODE (x);
switch (code)
{
case PC:
case CC0:
case CONST:
case CONST_INT:
case CONST_DOUBLE:
case CONST_VECTOR:
case SYMBOL_REF:
case LABEL_REF:
case ADDR_VEC:
case ADDR_DIFF_VEC:
return 1;
case MEM:
if (load_killed_in_block_p (BLOCK_FOR_INSN (insn),
INSN_CUID (insn), x, 0))
return 0;
else
return oprs_not_set_p (XEXP (x, 0), insn);
case REG:
return ! REGNO_REG_SET_P (reg_set_bitmap, REGNO (x));
default:
break;
}
for (i = GET_RTX_LENGTH (code) - 1, fmt = GET_RTX_FORMAT (code); i >= 0; i--)
{
if (fmt[i] == 'e')
{
/* If we are about to do the last recursive call
needed at this level, change it into iteration.
This function is called enough to be worth it. */
if (i == 0)
return oprs_not_set_p (XEXP (x, i), insn);
if (! oprs_not_set_p (XEXP (x, i), insn))
return 0;
}
else if (fmt[i] == 'E')
for (j = 0; j < XVECLEN (x, i); j++)
if (! oprs_not_set_p (XVECEXP (x, i, j), insn))
return 0;
}
return 1;
}
/* Mark things set by a CALL. */
static void
mark_call (rtx insn)
{
if (! CONST_OR_PURE_CALL_P (insn))
record_last_mem_set_info (insn);
}
/* Mark things set by a SET. */
static void
mark_set (rtx pat, rtx insn)
{
rtx dest = SET_DEST (pat);
while (GET_CODE (dest) == SUBREG
|| GET_CODE (dest) == ZERO_EXTRACT
|| GET_CODE (dest) == SIGN_EXTRACT
|| GET_CODE (dest) == STRICT_LOW_PART)
dest = XEXP (dest, 0);
if (GET_CODE (dest) == REG)
SET_REGNO_REG_SET (reg_set_bitmap, REGNO (dest));
else if (GET_CODE (dest) == MEM)
record_last_mem_set_info (insn);
if (GET_CODE (SET_SRC (pat)) == CALL)
mark_call (insn);
}
/* Record things set by a CLOBBER. */
static void
mark_clobber (rtx pat, rtx insn)
{
rtx clob = XEXP (pat, 0);
while (GET_CODE (clob) == SUBREG || GET_CODE (clob) == STRICT_LOW_PART)
clob = XEXP (clob, 0);
if (GET_CODE (clob) == REG)
SET_REGNO_REG_SET (reg_set_bitmap, REGNO (clob));
else
record_last_mem_set_info (insn);
}
/* Record things set by INSN.
This data is used by oprs_not_set_p. */
static void
mark_oprs_set (rtx insn)
{
rtx pat = PATTERN (insn);
int i;
if (GET_CODE (pat) == SET)
mark_set (pat, insn);
else if (GET_CODE (pat) == PARALLEL)
for (i = 0; i < XVECLEN (pat, 0); i++)
{
rtx x = XVECEXP (pat, 0, i);
if (GET_CODE (x) == SET)
mark_set (x, insn);
else if (GET_CODE (x) == CLOBBER)
mark_clobber (x, insn);
else if (GET_CODE (x) == CALL)
mark_call (insn);
}
else if (GET_CODE (pat) == CLOBBER)
mark_clobber (pat, insn);
else if (GET_CODE (pat) == CALL)
mark_call (insn);
}
/* Classic GCSE reaching definition support. */
/* Allocate reaching def variables. */
static void
alloc_rd_mem (int n_blocks, int n_insns)
{
rd_kill = sbitmap_vector_alloc (n_blocks, n_insns);
sbitmap_vector_zero (rd_kill, n_blocks);
rd_gen = sbitmap_vector_alloc (n_blocks, n_insns);
sbitmap_vector_zero (rd_gen, n_blocks);
reaching_defs = sbitmap_vector_alloc (n_blocks, n_insns);
sbitmap_vector_zero (reaching_defs, n_blocks);
rd_out = sbitmap_vector_alloc (n_blocks, n_insns);
sbitmap_vector_zero (rd_out, n_blocks);
}
/* Free reaching def variables. */
static void
free_rd_mem (void)
{
sbitmap_vector_free (rd_kill);
sbitmap_vector_free (rd_gen);
sbitmap_vector_free (reaching_defs);
sbitmap_vector_free (rd_out);
}
/* Add INSN to the kills of BB. REGNO, set in BB, is killed by INSN. */
static void
handle_rd_kill_set (rtx insn, int regno, basic_block bb)
{
struct reg_set *this_reg;
for (this_reg = reg_set_table[regno]; this_reg; this_reg = this_reg ->next)
if (BLOCK_NUM (this_reg->insn) != BLOCK_NUM (insn))
SET_BIT (rd_kill[bb->index], INSN_CUID (this_reg->insn));
}
/* Compute the set of kill's for reaching definitions. */
static void
compute_kill_rd (void)
{
int cuid;
unsigned int regno;
int i;
basic_block bb;
/* For each block
For each set bit in `gen' of the block (i.e each insn which
generates a definition in the block)
Call the reg set by the insn corresponding to that bit regx
Look at the linked list starting at reg_set_table[regx]
For each setting of regx in the linked list, which is not in
this block
Set the bit in `kill' corresponding to that insn. */
FOR_EACH_BB (bb)
for (cuid = 0; cuid < max_cuid; cuid++)
if (TEST_BIT (rd_gen[bb->index], cuid))
{
rtx insn = CUID_INSN (cuid);
rtx pat = PATTERN (insn);
if (GET_CODE (insn) == CALL_INSN)
{
for (regno = 0; regno < FIRST_PSEUDO_REGISTER; regno++)
if (TEST_HARD_REG_BIT (regs_invalidated_by_call, regno))
handle_rd_kill_set (insn, regno, bb);
}
if (GET_CODE (pat) == PARALLEL)
{
for (i = XVECLEN (pat, 0) - 1; i >= 0; i--)
{
enum rtx_code code = GET_CODE (XVECEXP (pat, 0, i));
if ((code == SET || code == CLOBBER)
&& GET_CODE (XEXP (XVECEXP (pat, 0, i), 0)) == REG)
handle_rd_kill_set (insn,
REGNO (XEXP (XVECEXP (pat, 0, i), 0)),
bb);
}
}
else if (GET_CODE (pat) == SET && GET_CODE (SET_DEST (pat)) == REG)
/* Each setting of this register outside of this block
must be marked in the set of kills in this block. */
handle_rd_kill_set (insn, REGNO (SET_DEST (pat)), bb);
}
}
/* Compute the reaching definitions as in
Compilers Principles, Techniques, and Tools. Aho, Sethi, Ullman,
Chapter 10. It is the same algorithm as used for computing available
expressions but applied to the gens and kills of reaching definitions. */
static void
compute_rd (void)
{
int changed, passes;
basic_block bb;
FOR_EACH_BB (bb)
sbitmap_copy (rd_out[bb->index] /*dst*/, rd_gen[bb->index] /*src*/);
passes = 0;
changed = 1;
while (changed)
{
changed = 0;
FOR_EACH_BB (bb)
{
sbitmap_union_of_preds (reaching_defs[bb->index], rd_out, bb->index);
changed |= sbitmap_union_of_diff_cg (rd_out[bb->index], rd_gen[bb->index],
reaching_defs[bb->index], rd_kill[bb->index]);
}
passes++;
}
if (gcse_file)
fprintf (gcse_file, "reaching def computation: %d passes\n", passes);
}
/* Classic GCSE available expression support. */
/* Allocate memory for available expression computation. */
static void
alloc_avail_expr_mem (int n_blocks, int n_exprs)
{
ae_kill = sbitmap_vector_alloc (n_blocks, n_exprs);
sbitmap_vector_zero (ae_kill, n_blocks);
ae_gen = sbitmap_vector_alloc (n_blocks, n_exprs);
sbitmap_vector_zero (ae_gen, n_blocks);
ae_in = sbitmap_vector_alloc (n_blocks, n_exprs);
sbitmap_vector_zero (ae_in, n_blocks);
ae_out = sbitmap_vector_alloc (n_blocks, n_exprs);
sbitmap_vector_zero (ae_out, n_blocks);
}
static void
free_avail_expr_mem (void)
{
sbitmap_vector_free (ae_kill);
sbitmap_vector_free (ae_gen);
sbitmap_vector_free (ae_in);
sbitmap_vector_free (ae_out);
}
/* Compute the set of available expressions generated in each basic block. */
static void
compute_ae_gen (struct hash_table *expr_hash_table)
{
unsigned int i;
struct expr *expr;
struct occr *occr;
/* For each recorded occurrence of each expression, set ae_gen[bb][expr].
This is all we have to do because an expression is not recorded if it
is not available, and the only expressions we want to work with are the
ones that are recorded. */
for (i = 0; i < expr_hash_table->size; i++)
for (expr = expr_hash_table->table[i]; expr != 0; expr = expr->next_same_hash)
for (occr = expr->avail_occr; occr != 0; occr = occr->next)
SET_BIT (ae_gen[BLOCK_NUM (occr->insn)], expr->bitmap_index);
}
/* Return nonzero if expression X is killed in BB. */
static int
expr_killed_p (rtx x, basic_block bb)
{
int i, j;
enum rtx_code code;
const char *fmt;
if (x == 0)
return 1;
code = GET_CODE (x);
switch (code)
{
case REG:
return TEST_BIT (reg_set_in_block[bb->index], REGNO (x));
case MEM:
if (load_killed_in_block_p (bb, get_max_uid () + 1, x, 0))
return 1;
else
return expr_killed_p (XEXP (x, 0), bb);
case PC:
case CC0: /*FIXME*/
case CONST:
case CONST_INT:
case CONST_DOUBLE:
case CONST_VECTOR:
case SYMBOL_REF:
case LABEL_REF:
case ADDR_VEC:
case ADDR_DIFF_VEC:
return 0;
default:
break;
}
for (i = GET_RTX_LENGTH (code) - 1, fmt = GET_RTX_FORMAT (code); i >= 0; i--)
{
if (fmt[i] == 'e')
{
/* If we are about to do the last recursive call
needed at this level, change it into iteration.
This function is called enough to be worth it. */
if (i == 0)
return expr_killed_p (XEXP (x, i), bb);
else if (expr_killed_p (XEXP (x, i), bb))
return 1;
}
else if (fmt[i] == 'E')
for (j = 0; j < XVECLEN (x, i); j++)
if (expr_killed_p (XVECEXP (x, i, j), bb))
return 1;
}
return 0;
}
/* Compute the set of available expressions killed in each basic block. */
static void
compute_ae_kill (sbitmap *ae_gen, sbitmap *ae_kill,
struct hash_table *expr_hash_table)
{
basic_block bb;
unsigned int i;
struct expr *expr;
FOR_EACH_BB (bb)
for (i = 0; i < expr_hash_table->size; i++)
for (expr = expr_hash_table->table[i]; expr; expr = expr->next_same_hash)
{
/* Skip EXPR if generated in this block. */
if (TEST_BIT (ae_gen[bb->index], expr->bitmap_index))
continue;
if (expr_killed_p (expr->expr, bb))
SET_BIT (ae_kill[bb->index], expr->bitmap_index);
}
}
/* Actually perform the Classic GCSE optimizations. */
/* Return nonzero if occurrence OCCR of expression EXPR reaches block BB.
CHECK_SELF_LOOP is nonzero if we should consider a block reaching itself
as a positive reach. We want to do this when there are two computations
of the expression in the block.
VISITED is a pointer to a working buffer for tracking which BB's have
been visited. It is NULL for the top-level call.
We treat reaching expressions that go through blocks containing the same
reaching expression as "not reaching". E.g. if EXPR is generated in blocks
2 and 3, INSN is in block 4, and 2->3->4, we treat the expression in block
2 as not reaching. The intent is to improve the probability of finding
only one reaching expression and to reduce register lifetimes by picking
the closest such expression. */
static int
expr_reaches_here_p_work (struct occr *occr, struct expr *expr,
basic_block bb, int check_self_loop, char *visited)
{
edge pred;
for (pred = bb->pred; pred != NULL; pred = pred->pred_next)
{
basic_block pred_bb = pred->src;
if (visited[pred_bb->index])
/* This predecessor has already been visited. Nothing to do. */
;
else if (pred_bb == bb)
{
/* BB loops on itself. */
if (check_self_loop
&& TEST_BIT (ae_gen[pred_bb->index], expr->bitmap_index)
&& BLOCK_NUM (occr->insn) == pred_bb->index)
return 1;
visited[pred_bb->index] = 1;
}
/* Ignore this predecessor if it kills the expression. */
else if (TEST_BIT (ae_kill[pred_bb->index], expr->bitmap_index))
visited[pred_bb->index] = 1;
/* Does this predecessor generate this expression? */
else if (TEST_BIT (ae_gen[pred_bb->index], expr->bitmap_index))
{
/* Is this the occurrence we're looking for?
Note that there's only one generating occurrence per block
so we just need to check the block number. */
if (BLOCK_NUM (occr->insn) == pred_bb->index)
return 1;
visited[pred_bb->index] = 1;
}
/* Neither gen nor kill. */
else
{
visited[pred_bb->index] = 1;
if (expr_reaches_here_p_work (occr, expr, pred_bb, check_self_loop,
visited))
return 1;
}
}
/* All paths have been checked. */
return 0;
}
/* This wrapper for expr_reaches_here_p_work() is to ensure that any
memory allocated for that function is returned. */
static int
expr_reaches_here_p (struct occr *occr, struct expr *expr, basic_block bb,
int check_self_loop)
{
int rval;
char *visited = xcalloc (last_basic_block, 1);
rval = expr_reaches_here_p_work (occr, expr, bb, check_self_loop, visited);
free (visited);
return rval;
}
/* Return the instruction that computes EXPR that reaches INSN's basic block.
If there is more than one such instruction, return NULL.
Called only by handle_avail_expr. */
static rtx
computing_insn (struct expr *expr, rtx insn)
{
basic_block bb = BLOCK_FOR_INSN (insn);
if (expr->avail_occr->next == NULL)
{
if (BLOCK_FOR_INSN (expr->avail_occr->insn) == bb)
/* The available expression is actually itself
(i.e. a loop in the flow graph) so do nothing. */
return NULL;
/* (FIXME) Case that we found a pattern that was created by
a substitution that took place. */
return expr->avail_occr->insn;
}
else
{
/* Pattern is computed more than once.
Search backwards from this insn to see how many of these
computations actually reach this insn. */
struct occr *occr;
rtx insn_computes_expr = NULL;
int can_reach = 0;
for (occr = expr->avail_occr; occr != NULL; occr = occr->next)
{
if (BLOCK_FOR_INSN (occr->insn) == bb)
{
/* The expression is generated in this block.
The only time we care about this is when the expression
is generated later in the block [and thus there's a loop].
We let the normal cse pass handle the other cases. */
if (INSN_CUID (insn) < INSN_CUID (occr->insn)
&& expr_reaches_here_p (occ