OpenCores
URL https://opencores.org/ocsvn/openrisc/openrisc/trunk

Subversion Repositories openrisc

[/] [openrisc/] [trunk/] [gnu-dev/] [or1k-gcc/] [gcc/] [doc/] [tree-ssa.texi] - Blame information for rev 711

Details | Compare with Previous | View Log

Line No. Rev Author Line
1 711 jeremybenn
@c Copyright (c) 2004, 2005, 2007, 2008, 2010
2
@c Free Software Foundation, Inc.
3
@c This is part of the GCC manual.
4
@c For copying conditions, see the file gcc.texi.
5
 
6
@c ---------------------------------------------------------------------
7
@c Tree SSA
8
@c ---------------------------------------------------------------------
9
 
10
@node Tree SSA
11
@chapter Analysis and Optimization of GIMPLE tuples
12
@cindex Tree SSA
13
@cindex Optimization infrastructure for GIMPLE
14
 
15
GCC uses three main intermediate languages to represent the program
16
during compilation: GENERIC, GIMPLE and RTL@.  GENERIC is a
17
language-independent representation generated by each front end.  It
18
is used to serve as an interface between the parser and optimizer.
19
GENERIC is a common representation that is able to represent programs
20
written in all the languages supported by GCC@.
21
 
22
GIMPLE and RTL are used to optimize the program.  GIMPLE is used for
23
target and language independent optimizations (e.g., inlining,
24
constant propagation, tail call elimination, redundancy elimination,
25
etc).  Much like GENERIC, GIMPLE is a language independent, tree based
26
representation.  However, it differs from GENERIC in that the GIMPLE
27
grammar is more restrictive: expressions contain no more than 3
28
operands (except function calls), it has no control flow structures
29
and expressions with side-effects are only allowed on the right hand
30
side of assignments.  See the chapter describing GENERIC and GIMPLE
31
for more details.
32
 
33
This chapter describes the data structures and functions used in the
34
GIMPLE optimizers (also known as ``tree optimizers'' or ``middle
35
end'').  In particular, it focuses on all the macros, data structures,
36
functions and programming constructs needed to implement optimization
37
passes for GIMPLE@.
38
 
39
@menu
40
* Annotations::         Attributes for variables.
41
* SSA Operands::        SSA names referenced by GIMPLE statements.
42
* SSA::                 Static Single Assignment representation.
43
* Alias analysis::      Representing aliased loads and stores.
44
* Memory model::        Memory model used by the middle-end.
45
@end menu
46
 
47
@node Annotations
48
@section Annotations
49
@cindex annotations
50
 
51
The optimizers need to associate attributes with variables during the
52
optimization process.  For instance, we need to know whether a
53
variable has aliases.  All these attributes are stored in data
54
structures called annotations which are then linked to the field
55
@code{ann} in @code{struct tree_common}.
56
 
57
Presently, we define annotations for variables (@code{var_ann_t}).
58
Annotations are defined and documented in @file{tree-flow.h}.
59
 
60
 
61
@node SSA Operands
62
@section SSA Operands
63
@cindex operands
64
@cindex virtual operands
65
@cindex real operands
66
@findex update_stmt
67
 
68
Almost every GIMPLE statement will contain a reference to a variable
69
or memory location.  Since statements come in different shapes and
70
sizes, their operands are going to be located at various spots inside
71
the statement's tree.  To facilitate access to the statement's
72
operands, they are organized into lists associated inside each
73
statement's annotation.  Each element in an operand list is a pointer
74
to a @code{VAR_DECL}, @code{PARM_DECL} or @code{SSA_NAME} tree node.
75
This provides a very convenient way of examining and replacing
76
operands.
77
 
78
Data flow analysis and optimization is done on all tree nodes
79
representing variables.  Any node for which @code{SSA_VAR_P} returns
80
nonzero is considered when scanning statement operands.  However, not
81
all @code{SSA_VAR_P} variables are processed in the same way.  For the
82
purposes of optimization, we need to distinguish between references to
83
local scalar variables and references to globals, statics, structures,
84
arrays, aliased variables, etc.  The reason is simple, the compiler
85
can gather complete data flow information for a local scalar.  On the
86
other hand, a global variable may be modified by a function call, it
87
may not be possible to keep track of all the elements of an array or
88
the fields of a structure, etc.
89
 
90
The operand scanner gathers two kinds of operands: @dfn{real} and
91
@dfn{virtual}.  An operand for which @code{is_gimple_reg} returns true
92
is considered real, otherwise it is a virtual operand.  We also
93
distinguish between uses and definitions.  An operand is used if its
94
value is loaded by the statement (e.g., the operand at the RHS of an
95
assignment).  If the statement assigns a new value to the operand, the
96
operand is considered a definition (e.g., the operand at the LHS of
97
an assignment).
98
 
99
Virtual and real operands also have very different data flow
100
properties.  Real operands are unambiguous references to the
101
full object that they represent.  For instance, given
102
 
103
@smallexample
104
@{
105
  int a, b;
106
  a = b
107
@}
108
@end smallexample
109
 
110
Since @code{a} and @code{b} are non-aliased locals, the statement
111
@code{a = b} will have one real definition and one real use because
112
variable @code{a} is completely modified with the contents of
113
variable @code{b}.  Real definition are also known as @dfn{killing
114
definitions}.  Similarly, the use of @code{b} reads all its bits.
115
 
116
In contrast, virtual operands are used with variables that can have
117
a partial or ambiguous reference.  This includes structures, arrays,
118
globals, and aliased variables.  In these cases, we have two types of
119
definitions.  For globals, structures, and arrays, we can determine from
120
a statement whether a variable of these types has a killing definition.
121
If the variable does, then the statement is marked as having a
122
@dfn{must definition} of that variable.  However, if a statement is only
123
defining a part of the variable (i.e.@: a field in a structure), or if we
124
know that a statement might define the variable but we cannot say for sure,
125
then we mark that statement as having a @dfn{may definition}.  For
126
instance, given
127
 
128
@smallexample
129
@{
130
  int a, b, *p;
131
 
132
  if (@dots{})
133
    p = &a;
134
  else
135
    p = &b;
136
  *p = 5;
137
  return *p;
138
@}
139
@end smallexample
140
 
141
The assignment @code{*p = 5} may be a definition of @code{a} or
142
@code{b}.  If we cannot determine statically where @code{p} is
143
pointing to at the time of the store operation, we create virtual
144
definitions to mark that statement as a potential definition site for
145
@code{a} and @code{b}.  Memory loads are similarly marked with virtual
146
use operands.  Virtual operands are shown in tree dumps right before
147
the statement that contains them.  To request a tree dump with virtual
148
operands, use the @option{-vops} option to @option{-fdump-tree}:
149
 
150
@smallexample
151
@{
152
  int a, b, *p;
153
 
154
  if (@dots{})
155
    p = &a;
156
  else
157
    p = &b;
158
  # a = VDEF <a>
159
  # b = VDEF <b>
160
  *p = 5;
161
 
162
  # VUSE <a>
163
  # VUSE <b>
164
  return *p;
165
@}
166
@end smallexample
167
 
168
Notice that @code{VDEF} operands have two copies of the referenced
169
variable.  This indicates that this is not a killing definition of
170
that variable.  In this case we refer to it as a @dfn{may definition}
171
or @dfn{aliased store}.  The presence of the second copy of the
172
variable in the @code{VDEF} operand will become important when the
173
function is converted into SSA form.  This will be used to link all
174
the non-killing definitions to prevent optimizations from making
175
incorrect assumptions about them.
176
 
177
Operands are updated as soon as the statement is finished via a call
178
to @code{update_stmt}.  If statement elements are changed via
179
@code{SET_USE} or @code{SET_DEF}, then no further action is required
180
(i.e., those macros take care of updating the statement).  If changes
181
are made by manipulating the statement's tree directly, then a call
182
must be made to @code{update_stmt} when complete.  Calling one of the
183
@code{bsi_insert} routines or @code{bsi_replace} performs an implicit
184
call to @code{update_stmt}.
185
 
186
@subsection Operand Iterators And Access Routines
187
@cindex Operand Iterators
188
@cindex Operand Access Routines
189
 
190
Operands are collected by @file{tree-ssa-operands.c}.  They are stored
191
inside each statement's annotation and can be accessed through either the
192
operand iterators or an access routine.
193
 
194
The following access routines are available for examining operands:
195
 
196
@enumerate
197
@item @code{SINGLE_SSA_@{USE,DEF,TREE@}_OPERAND}: These accessors will return
198
NULL unless there is exactly one operand matching the specified flags.  If
199
there is exactly one operand, the operand is returned as either a @code{tree},
200
@code{def_operand_p}, or @code{use_operand_p}.
201
 
202
@smallexample
203
tree t = SINGLE_SSA_TREE_OPERAND (stmt, flags);
204
use_operand_p u = SINGLE_SSA_USE_OPERAND (stmt, SSA_ALL_VIRTUAL_USES);
205
def_operand_p d = SINGLE_SSA_DEF_OPERAND (stmt, SSA_OP_ALL_DEFS);
206
@end smallexample
207
 
208
@item @code{ZERO_SSA_OPERANDS}: This macro returns true if there are no
209
operands matching the specified flags.
210
 
211
@smallexample
212
if (ZERO_SSA_OPERANDS (stmt, SSA_OP_ALL_VIRTUALS))
213
  return;
214
@end smallexample
215
 
216
@item @code{NUM_SSA_OPERANDS}: This macro Returns the number of operands
217
matching 'flags'.  This actually executes a loop to perform the count, so
218
only use this if it is really needed.
219
 
220
@smallexample
221
int count = NUM_SSA_OPERANDS (stmt, flags)
222
@end smallexample
223
@end enumerate
224
 
225
 
226
If you wish to iterate over some or all operands, use the
227
@code{FOR_EACH_SSA_@{USE,DEF,TREE@}_OPERAND} iterator.  For example, to print
228
all the operands for a statement:
229
 
230
@smallexample
231
void
232
print_ops (tree stmt)
233
@{
234
  ssa_op_iter;
235
  tree var;
236
 
237
  FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_ALL_OPERANDS)
238
    print_generic_expr (stderr, var, TDF_SLIM);
239
@}
240
@end smallexample
241
 
242
 
243
How to choose the appropriate iterator:
244
 
245
@enumerate
246
@item Determine whether you are need to see the operand pointers, or just the
247
trees, and choose the appropriate macro:
248
 
249
@smallexample
250
Need            Macro:
251
----            -------
252
use_operand_p   FOR_EACH_SSA_USE_OPERAND
253
def_operand_p   FOR_EACH_SSA_DEF_OPERAND
254
tree            FOR_EACH_SSA_TREE_OPERAND
255
@end smallexample
256
 
257
@item You need to declare a variable of the type you are interested
258
in, and an ssa_op_iter structure which serves as the loop controlling
259
variable.
260
 
261
@item Determine which operands you wish to use, and specify the flags of
262
those you are interested in.  They are documented in
263
@file{tree-ssa-operands.h}:
264
 
265
@smallexample
266
#define SSA_OP_USE              0x01    /* @r{Real USE operands.}  */
267
#define SSA_OP_DEF              0x02    /* @r{Real DEF operands.}  */
268
#define SSA_OP_VUSE             0x04    /* @r{VUSE operands.}  */
269
#define SSA_OP_VMAYUSE          0x08    /* @r{USE portion of VDEFS.}  */
270
#define SSA_OP_VDEF             0x10    /* @r{DEF portion of VDEFS.}  */
271
 
272
/* @r{These are commonly grouped operand flags.}  */
273
#define SSA_OP_VIRTUAL_USES     (SSA_OP_VUSE | SSA_OP_VMAYUSE)
274
#define SSA_OP_VIRTUAL_DEFS     (SSA_OP_VDEF)
275
#define SSA_OP_ALL_USES         (SSA_OP_VIRTUAL_USES | SSA_OP_USE)
276
#define SSA_OP_ALL_DEFS         (SSA_OP_VIRTUAL_DEFS | SSA_OP_DEF)
277
#define SSA_OP_ALL_OPERANDS     (SSA_OP_ALL_USES | SSA_OP_ALL_DEFS)
278
@end smallexample
279
@end enumerate
280
 
281
So if you want to look at the use pointers for all the @code{USE} and
282
@code{VUSE} operands, you would do something like:
283
 
284
@smallexample
285
  use_operand_p use_p;
286
  ssa_op_iter iter;
287
 
288
  FOR_EACH_SSA_USE_OPERAND (use_p, stmt, iter, (SSA_OP_USE | SSA_OP_VUSE))
289
    @{
290
      process_use_ptr (use_p);
291
    @}
292
@end smallexample
293
 
294
The @code{TREE} macro is basically the same as the @code{USE} and
295
@code{DEF} macros, only with the use or def dereferenced via
296
@code{USE_FROM_PTR (use_p)} and @code{DEF_FROM_PTR (def_p)}.  Since we
297
aren't using operand pointers, use and defs flags can be mixed.
298
 
299
@smallexample
300
  tree var;
301
  ssa_op_iter iter;
302
 
303
  FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_VUSE)
304
    @{
305
       print_generic_expr (stderr, var, TDF_SLIM);
306
    @}
307
@end smallexample
308
 
309
@code{VDEF}s are broken into two flags, one for the
310
@code{DEF} portion (@code{SSA_OP_VDEF}) and one for the USE portion
311
(@code{SSA_OP_VMAYUSE}).  If all you want to look at are the
312
@code{VDEF}s together, there is a fourth iterator macro for this,
313
which returns both a def_operand_p and a use_operand_p for each
314
@code{VDEF} in the statement.  Note that you don't need any flags for
315
this one.
316
 
317
@smallexample
318
  use_operand_p use_p;
319
  def_operand_p def_p;
320
  ssa_op_iter iter;
321
 
322
  FOR_EACH_SSA_MAYDEF_OPERAND (def_p, use_p, stmt, iter)
323
    @{
324
      my_code;
325
    @}
326
@end smallexample
327
 
328
There are many examples in the code as well, as well as the
329
documentation in @file{tree-ssa-operands.h}.
330
 
331
There are also a couple of variants on the stmt iterators regarding PHI
332
nodes.
333
 
334
@code{FOR_EACH_PHI_ARG} Works exactly like
335
@code{FOR_EACH_SSA_USE_OPERAND}, except it works over @code{PHI} arguments
336
instead of statement operands.
337
 
338
@smallexample
339
/* Look at every virtual PHI use.  */
340
FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_VIRTUAL_USES)
341
@{
342
   my_code;
343
@}
344
 
345
/* Look at every real PHI use.  */
346
FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_USES)
347
  my_code;
348
 
349
/* Look at every PHI use.  */
350
FOR_EACH_PHI_ARG (use_p, phi_stmt, iter, SSA_OP_ALL_USES)
351
  my_code;
352
@end smallexample
353
 
354
@code{FOR_EACH_PHI_OR_STMT_@{USE,DEF@}} works exactly like
355
@code{FOR_EACH_SSA_@{USE,DEF@}_OPERAND}, except it will function on
356
either a statement or a @code{PHI} node.  These should be used when it is
357
appropriate but they are not quite as efficient as the individual
358
@code{FOR_EACH_PHI} and @code{FOR_EACH_SSA} routines.
359
 
360
@smallexample
361
FOR_EACH_PHI_OR_STMT_USE (use_operand_p, stmt, iter, flags)
362
  @{
363
     my_code;
364
  @}
365
 
366
FOR_EACH_PHI_OR_STMT_DEF (def_operand_p, phi, iter, flags)
367
  @{
368
     my_code;
369
  @}
370
@end smallexample
371
 
372
@subsection Immediate Uses
373
@cindex Immediate Uses
374
 
375
Immediate use information is now always available.  Using the immediate use
376
iterators, you may examine every use of any @code{SSA_NAME}. For instance,
377
to change each use of @code{ssa_var} to @code{ssa_var2} and call fold_stmt on
378
each stmt after that is done:
379
 
380
@smallexample
381
  use_operand_p imm_use_p;
382
  imm_use_iterator iterator;
383
  tree ssa_var, stmt;
384
 
385
 
386
  FOR_EACH_IMM_USE_STMT (stmt, iterator, ssa_var)
387
    @{
388
      FOR_EACH_IMM_USE_ON_STMT (imm_use_p, iterator)
389
        SET_USE (imm_use_p, ssa_var_2);
390
      fold_stmt (stmt);
391
    @}
392
@end smallexample
393
 
394
There are 2 iterators which can be used. @code{FOR_EACH_IMM_USE_FAST} is
395
used when the immediate uses are not changed, i.e., you are looking at the
396
uses, but not setting them.
397
 
398
If they do get changed, then care must be taken that things are not changed
399
under the iterators, so use the @code{FOR_EACH_IMM_USE_STMT} and
400
@code{FOR_EACH_IMM_USE_ON_STMT} iterators.  They attempt to preserve the
401
sanity of the use list by moving all the uses for a statement into
402
a controlled position, and then iterating over those uses.  Then the
403
optimization can manipulate the stmt when all the uses have been
404
processed.  This is a little slower than the FAST version since it adds a
405
placeholder element and must sort through the list a bit for each statement.
406
This placeholder element must be also be removed if the loop is
407
terminated early.  The macro @code{BREAK_FROM_IMM_USE_SAFE} is provided
408
to do this :
409
 
410
@smallexample
411
  FOR_EACH_IMM_USE_STMT (stmt, iterator, ssa_var)
412
    @{
413
      if (stmt == last_stmt)
414
        BREAK_FROM_SAFE_IMM_USE (iter);
415
 
416
      FOR_EACH_IMM_USE_ON_STMT (imm_use_p, iterator)
417
        SET_USE (imm_use_p, ssa_var_2);
418
      fold_stmt (stmt);
419
    @}
420
@end smallexample
421
 
422
There are checks in @code{verify_ssa} which verify that the immediate use list
423
is up to date, as well as checking that an optimization didn't break from the
424
loop without using this macro.  It is safe to simply 'break'; from a
425
@code{FOR_EACH_IMM_USE_FAST} traverse.
426
 
427
Some useful functions and macros:
428
@enumerate
429
@item  @code{has_zero_uses (ssa_var)} : Returns true if there are no uses of
430
@code{ssa_var}.
431
@item   @code{has_single_use (ssa_var)} : Returns true if there is only a
432
single use of @code{ssa_var}.
433
@item   @code{single_imm_use (ssa_var, use_operand_p *ptr, tree *stmt)} :
434
Returns true if there is only a single use of @code{ssa_var}, and also returns
435
the use pointer and statement it occurs in, in the second and third parameters.
436
@item   @code{num_imm_uses (ssa_var)} : Returns the number of immediate uses of
437
@code{ssa_var}. It is better not to use this if possible since it simply
438
utilizes a loop to count the uses.
439
@item  @code{PHI_ARG_INDEX_FROM_USE (use_p)} : Given a use within a @code{PHI}
440
node, return the index number for the use.  An assert is triggered if the use
441
isn't located in a @code{PHI} node.
442
@item  @code{USE_STMT (use_p)} : Return the statement a use occurs in.
443
@end enumerate
444
 
445
Note that uses are not put into an immediate use list until their statement is
446
actually inserted into the instruction stream via a @code{bsi_*} routine.
447
 
448
It is also still possible to utilize lazy updating of statements, but this
449
should be used only when absolutely required.  Both alias analysis and the
450
dominator optimizations currently do this.
451
 
452
When lazy updating is being used, the immediate use information is out of date
453
and cannot be used reliably.  Lazy updating is achieved by simply marking
454
statements modified via calls to @code{mark_stmt_modified} instead of
455
@code{update_stmt}.  When lazy updating is no longer required, all the
456
modified statements must have @code{update_stmt} called in order to bring them
457
up to date.  This must be done before the optimization is finished, or
458
@code{verify_ssa} will trigger an abort.
459
 
460
This is done with a simple loop over the instruction stream:
461
@smallexample
462
  block_stmt_iterator bsi;
463
  basic_block bb;
464
  FOR_EACH_BB (bb)
465
    @{
466
      for (bsi = bsi_start (bb); !bsi_end_p (bsi); bsi_next (&bsi))
467
        update_stmt_if_modified (bsi_stmt (bsi));
468
    @}
469
@end smallexample
470
 
471
@node SSA
472
@section Static Single Assignment
473
@cindex SSA
474
@cindex static single assignment
475
 
476
Most of the tree optimizers rely on the data flow information provided
477
by the Static Single Assignment (SSA) form.  We implement the SSA form
478
as described in @cite{R. Cytron, J. Ferrante, B. Rosen, M. Wegman, and
479
K. Zadeck.  Efficiently Computing Static Single Assignment Form and the
480
Control Dependence Graph.  ACM Transactions on Programming Languages
481
and Systems, 13(4):451-490, October 1991}.
482
 
483
The SSA form is based on the premise that program variables are
484
assigned in exactly one location in the program.  Multiple assignments
485
to the same variable create new versions of that variable.  Naturally,
486
actual programs are seldom in SSA form initially because variables
487
tend to be assigned multiple times.  The compiler modifies the program
488
representation so that every time a variable is assigned in the code,
489
a new version of the variable is created.  Different versions of the
490
same variable are distinguished by subscripting the variable name with
491
its version number.  Variables used in the right-hand side of
492
expressions are renamed so that their version number matches that of
493
the most recent assignment.
494
 
495
We represent variable versions using @code{SSA_NAME} nodes.  The
496
renaming process in @file{tree-ssa.c} wraps every real and
497
virtual operand with an @code{SSA_NAME} node which contains
498
the version number and the statement that created the
499
@code{SSA_NAME}.  Only definitions and virtual definitions may
500
create new @code{SSA_NAME} nodes.
501
 
502
@cindex PHI nodes
503
Sometimes, flow of control makes it impossible to determine the
504
most recent version of a variable.  In these cases, the compiler
505
inserts an artificial definition for that variable called
506
@dfn{PHI function} or @dfn{PHI node}.  This new definition merges
507
all the incoming versions of the variable to create a new name
508
for it.  For instance,
509
 
510
@smallexample
511
if (@dots{})
512
  a_1 = 5;
513
else if (@dots{})
514
  a_2 = 2;
515
else
516
  a_3 = 13;
517
 
518
# a_4 = PHI <a_1, a_2, a_3>
519
return a_4;
520
@end smallexample
521
 
522
Since it is not possible to determine which of the three branches
523
will be taken at runtime, we don't know which of @code{a_1},
524
@code{a_2} or @code{a_3} to use at the return statement.  So, the
525
SSA renamer creates a new version @code{a_4} which is assigned
526
the result of ``merging'' @code{a_1}, @code{a_2} and @code{a_3}.
527
Hence, PHI nodes mean ``one of these operands.  I don't know
528
which''.
529
 
530
The following macros can be used to examine PHI nodes
531
 
532
@defmac PHI_RESULT (@var{phi})
533
Returns the @code{SSA_NAME} created by PHI node @var{phi} (i.e.,
534
@var{phi}'s LHS)@.
535
@end defmac
536
 
537
@defmac PHI_NUM_ARGS (@var{phi})
538
Returns the number of arguments in @var{phi}.  This number is exactly
539
the number of incoming edges to the basic block holding @var{phi}@.
540
@end defmac
541
 
542
@defmac PHI_ARG_ELT (@var{phi}, @var{i})
543
Returns a tuple representing the @var{i}th argument of @var{phi}@.
544
Each element of this tuple contains an @code{SSA_NAME} @var{var} and
545
the incoming edge through which @var{var} flows.
546
@end defmac
547
 
548
@defmac PHI_ARG_EDGE (@var{phi}, @var{i})
549
Returns the incoming edge for the @var{i}th argument of @var{phi}.
550
@end defmac
551
 
552
@defmac PHI_ARG_DEF (@var{phi}, @var{i})
553
Returns the @code{SSA_NAME} for the @var{i}th argument of @var{phi}.
554
@end defmac
555
 
556
 
557
@subsection Preserving the SSA form
558
@findex update_ssa
559
@cindex preserving SSA form
560
Some optimization passes make changes to the function that
561
invalidate the SSA property.  This can happen when a pass has
562
added new symbols or changed the program so that variables that
563
were previously aliased aren't anymore.  Whenever something like this
564
happens, the affected symbols must be renamed into SSA form again.
565
Transformations that emit new code or replicate existing statements
566
will also need to update the SSA form@.
567
 
568
Since GCC implements two different SSA forms for register and virtual
569
variables, keeping the SSA form up to date depends on whether you are
570
updating register or virtual names.  In both cases, the general idea
571
behind incremental SSA updates is similar: when new SSA names are
572
created, they typically are meant to replace other existing names in
573
the program@.
574
 
575
For instance, given the following code:
576
 
577
@smallexample
578
     1  L0:
579
     2  x_1 = PHI (0, x_5)
580
     3  if (x_1 < 10)
581
     4    if (x_1 > 7)
582
     5      y_2 = 0
583
     6    else
584
     7      y_3 = x_1 + x_7
585
     8    endif
586
     9    x_5 = x_1 + 1
587
     10   goto L0;
588
     11 endif
589
@end smallexample
590
 
591
Suppose that we insert new names @code{x_10} and @code{x_11} (lines
592
@code{4} and @code{8})@.
593
 
594
@smallexample
595
     1  L0:
596
     2  x_1 = PHI (0, x_5)
597
     3  if (x_1 < 10)
598
     4    x_10 = @dots{}
599
     5    if (x_1 > 7)
600
     6      y_2 = 0
601
     7    else
602
     8      x_11 = @dots{}
603
     9      y_3 = x_1 + x_7
604
     10   endif
605
     11   x_5 = x_1 + 1
606
     12   goto L0;
607
     13 endif
608
@end smallexample
609
 
610
We want to replace all the uses of @code{x_1} with the new definitions
611
of @code{x_10} and @code{x_11}.  Note that the only uses that should
612
be replaced are those at lines @code{5}, @code{9} and @code{11}.
613
Also, the use of @code{x_7} at line @code{9} should @emph{not} be
614
replaced (this is why we cannot just mark symbol @code{x} for
615
renaming)@.
616
 
617
Additionally, we may need to insert a PHI node at line @code{11}
618
because that is a merge point for @code{x_10} and @code{x_11}.  So the
619
use of @code{x_1} at line @code{11} will be replaced with the new PHI
620
node.  The insertion of PHI nodes is optional.  They are not strictly
621
necessary to preserve the SSA form, and depending on what the caller
622
inserted, they may not even be useful for the optimizers@.
623
 
624
Updating the SSA form is a two step process.  First, the pass has to
625
identify which names need to be updated and/or which symbols need to
626
be renamed into SSA form for the first time.  When new names are
627
introduced to replace existing names in the program, the mapping
628
between the old and the new names are registered by calling
629
@code{register_new_name_mapping} (note that if your pass creates new
630
code by duplicating basic blocks, the call to @code{tree_duplicate_bb}
631
will set up the necessary mappings automatically).  On the other hand,
632
if your pass exposes a new symbol that should be put in SSA form for
633
the first time, the new symbol should be registered with
634
@code{mark_sym_for_renaming}.
635
 
636
After the replacement mappings have been registered and new symbols
637
marked for renaming, a call to @code{update_ssa} makes the registered
638
changes.  This can be done with an explicit call or by creating
639
@code{TODO} flags in the @code{tree_opt_pass} structure for your pass.
640
There are several @code{TODO} flags that control the behavior of
641
@code{update_ssa}:
642
 
643
@itemize @bullet
644
@item @code{TODO_update_ssa}.  Update the SSA form inserting PHI nodes
645
for newly exposed symbols and virtual names marked for updating.
646
When updating real names, only insert PHI nodes for a real name
647
@code{O_j} in blocks reached by all the new and old definitions for
648
@code{O_j}.  If the iterated dominance frontier for @code{O_j}
649
is not pruned, we may end up inserting PHI nodes in blocks that
650
have one or more edges with no incoming definition for
651
@code{O_j}.  This would lead to uninitialized warnings for
652
@code{O_j}'s symbol@.
653
 
654
@item @code{TODO_update_ssa_no_phi}.  Update the SSA form without
655
inserting any new PHI nodes at all.  This is used by passes that
656
have either inserted all the PHI nodes themselves or passes that
657
need only to patch use-def and def-def chains for virtuals
658
(e.g., DCE)@.
659
 
660
 
661
@item @code{TODO_update_ssa_full_phi}.  Insert PHI nodes everywhere
662
they are needed.  No pruning of the IDF is done.  This is used
663
by passes that need the PHI nodes for @code{O_j} even if it
664
means that some arguments will come from the default definition
665
of @code{O_j}'s symbol (e.g., @code{pass_linear_transform})@.
666
 
667
WARNING: If you need to use this flag, chances are that your
668
pass may be doing something wrong.  Inserting PHI nodes for an
669
old name where not all edges carry a new replacement may lead to
670
silent codegen errors or spurious uninitialized warnings@.
671
 
672
@item @code{TODO_update_ssa_only_virtuals}.  Passes that update the
673
SSA form on their own may want to delegate the updating of
674
virtual names to the generic updater.  Since FUD chains are
675
easier to maintain, this simplifies the work they need to do.
676
NOTE: If this flag is used, any OLD->NEW mappings for real names
677
are explicitly destroyed and only the symbols marked for
678
renaming are processed@.
679
@end itemize
680
 
681
@subsection Preserving the virtual SSA form
682
@cindex preserving virtual SSA form
683
 
684
The virtual SSA form is harder to preserve than the non-virtual SSA form
685
mainly because the set of virtual operands for a statement may change at
686
what some would consider unexpected times.  In general, statement
687
modifications should be bracketed between calls to
688
@code{push_stmt_changes} and @code{pop_stmt_changes}.  For example,
689
 
690
@smallexample
691
    munge_stmt (tree stmt)
692
    @{
693
       push_stmt_changes (&stmt);
694
       @dots{} rewrite STMT @dots{}
695
       pop_stmt_changes (&stmt);
696
    @}
697
@end smallexample
698
 
699
The call to @code{push_stmt_changes} saves the current state of the
700
statement operands and the call to @code{pop_stmt_changes} compares
701
the saved state with the current one and does the appropriate symbol
702
marking for the SSA renamer.
703
 
704
It is possible to modify several statements at a time, provided that
705
@code{push_stmt_changes} and @code{pop_stmt_changes} are called in
706
LIFO order, as when processing a stack of statements.
707
 
708
Additionally, if the pass discovers that it did not need to make
709
changes to the statement after calling @code{push_stmt_changes}, it
710
can simply discard the topmost change buffer by calling
711
@code{discard_stmt_changes}.  This will avoid the expensive operand
712
re-scan operation and the buffer comparison that determines if symbols
713
need to be marked for renaming.
714
 
715
@subsection Examining @code{SSA_NAME} nodes
716
@cindex examining SSA_NAMEs
717
 
718
The following macros can be used to examine @code{SSA_NAME} nodes
719
 
720
@defmac SSA_NAME_DEF_STMT (@var{var})
721
Returns the statement @var{s} that creates the @code{SSA_NAME}
722
@var{var}.  If @var{s} is an empty statement (i.e., @code{IS_EMPTY_STMT
723
(@var{s})} returns @code{true}), it means that the first reference to
724
this variable is a USE or a VUSE@.
725
@end defmac
726
 
727
@defmac SSA_NAME_VERSION (@var{var})
728
Returns the version number of the @code{SSA_NAME} object @var{var}.
729
@end defmac
730
 
731
 
732
@subsection Walking use-def chains
733
 
734
@deftypefn {Tree SSA function} void walk_use_def_chains (@var{var}, @var{fn}, @var{data})
735
 
736
Walks use-def chains starting at the @code{SSA_NAME} node @var{var}.
737
Calls function @var{fn} at each reaching definition found.  Function
738
@var{FN} takes three arguments: @var{var}, its defining statement
739
(@var{def_stmt}) and a generic pointer to whatever state information
740
that @var{fn} may want to maintain (@var{data}).  Function @var{fn} is
741
able to stop the walk by returning @code{true}, otherwise in order to
742
continue the walk, @var{fn} should return @code{false}.
743
 
744
Note, that if @var{def_stmt} is a @code{PHI} node, the semantics are
745
slightly different.  For each argument @var{arg} of the PHI node, this
746
function will:
747
 
748
@enumerate
749
@item Walk the use-def chains for @var{arg}.
750
@item Call @code{FN (@var{arg}, @var{phi}, @var{data})}.
751
@end enumerate
752
 
753
Note how the first argument to @var{fn} is no longer the original
754
variable @var{var}, but the PHI argument currently being examined.
755
If @var{fn} wants to get at @var{var}, it should call
756
@code{PHI_RESULT} (@var{phi}).
757
@end deftypefn
758
 
759
@subsection Walking the dominator tree
760
 
761
@deftypefn {Tree SSA function} void walk_dominator_tree (@var{walk_data}, @var{bb})
762
 
763
This function walks the dominator tree for the current CFG calling a
764
set of callback functions defined in @var{struct dom_walk_data} in
765
@file{domwalk.h}.  The call back functions you need to define give you
766
hooks to execute custom code at various points during traversal:
767
 
768
@enumerate
769
@item Once to initialize any local data needed while processing
770
@var{bb} and its children.  This local data is pushed into an
771
internal stack which is automatically pushed and popped as the
772
walker traverses the dominator tree.
773
 
774
@item Once before traversing all the statements in the @var{bb}.
775
 
776
@item Once for every statement inside @var{bb}.
777
 
778
@item Once after traversing all the statements and before recursing
779
into @var{bb}'s dominator children.
780
 
781
@item It then recurses into all the dominator children of @var{bb}.
782
 
783
@item After recursing into all the dominator children of @var{bb} it
784
can, optionally, traverse every statement in @var{bb} again
785
(i.e., repeating steps 2 and 3).
786
 
787
@item Once after walking the statements in @var{bb} and @var{bb}'s
788
dominator children.  At this stage, the block local data stack
789
is popped.
790
@end enumerate
791
@end deftypefn
792
 
793
@node Alias analysis
794
@section Alias analysis
795
@cindex alias
796
@cindex flow-sensitive alias analysis
797
@cindex flow-insensitive alias analysis
798
 
799
Alias analysis in GIMPLE SSA form consists of two pieces.  First
800
the virtual SSA web ties conflicting memory accesses and provides
801
a SSA use-def chain and SSA immediate-use chains for walking
802
possibly dependent memory accesses.  Second an alias-oracle can
803
be queried to disambiguate explicit and implicit memory references.
804
 
805
@enumerate
806
@item Memory SSA form.
807
 
808
All statements that may use memory have exactly one accompanied use of
809
a virtual SSA name that represents the state of memory at the
810
given point in the IL.
811
 
812
All statements that may define memory have exactly one accompanied
813
definition of a virtual SSA name using the previous state of memory
814
and defining the new state of memory after the given point in the IL.
815
 
816
@smallexample
817
int i;
818
int foo (void)
819
@{
820
  # .MEM_3 = VDEF <.MEM_2(D)>
821
  i = 1;
822
  # VUSE <.MEM_3>
823
  return i;
824
@}
825
@end smallexample
826
 
827
The virtual SSA names in this case are @code{.MEM_2(D)} and
828
@code{.MEM_3}.  The store to the global variable @code{i}
829
defines @code{.MEM_3} invalidating @code{.MEM_2(D)}.  The
830
load from @code{i} uses that new state @code{.MEM_3}.
831
 
832
The virtual SSA web serves as constraints to SSA optimizers
833
preventing illegitimate code-motion and optimization.  It
834
also provides a way to walk related memory statements.
835
 
836
@item Points-to and escape analysis.
837
 
838
Points-to analysis builds a set of constraints from the GIMPLE
839
SSA IL representing all pointer operations and facts we do
840
or do not know about pointers.  Solving this set of constraints
841
yields a conservatively correct solution for each pointer
842
variable in the program (though we are only interested in
843
SSA name pointers) as to what it may possibly point to.
844
 
845
This points-to solution for a given SSA name pointer is stored
846
in the @code{pt_solution} sub-structure of the
847
@code{SSA_NAME_PTR_INFO} record.  The following accessor
848
functions are available:
849
 
850
@itemize @bullet
851
@item @code{pt_solution_includes}
852
@item @code{pt_solutions_intersect}
853
@end itemize
854
 
855
Points-to analysis also computes the solution for two special
856
set of pointers, @code{ESCAPED} and @code{CALLUSED}.  Those
857
represent all memory that has escaped the scope of analysis
858
or that is used by pure or nested const calls.
859
 
860
@item Type-based alias analysis
861
 
862
Type-based alias analysis is frontend dependent though generic
863
support is provided by the middle-end in @code{alias.c}.  TBAA
864
code is used by both tree optimizers and RTL optimizers.
865
 
866
Every language that wishes to perform language-specific alias analysis
867
should define a function that computes, given a @code{tree}
868
node, an alias set for the node.  Nodes in different alias sets are not
869
allowed to alias.  For an example, see the C front-end function
870
@code{c_get_alias_set}.
871
 
872
@item Tree alias-oracle
873
 
874
The tree alias-oracle provides means to disambiguate two memory
875
references and memory references against statements.  The following
876
queries are available:
877
 
878
@itemize @bullet
879
@item @code{refs_may_alias_p}
880
@item @code{ref_maybe_used_by_stmt_p}
881
@item @code{stmt_may_clobber_ref_p}
882
@end itemize
883
 
884
In addition to those two kind of statement walkers are available
885
walking statements related to a reference ref.
886
@code{walk_non_aliased_vuses} walks over dominating memory defining
887
statements and calls back if the statement does not clobber ref
888
providing the non-aliased VUSE.  The walk stops at
889
the first clobbering statement or if asked to.
890
@code{walk_aliased_vdefs} walks over dominating memory defining
891
statements and calls back on each statement clobbering ref
892
providing its aliasing VDEF.  The walk stops if asked to.
893
 
894
@end enumerate
895
 
896
 
897
@node Memory model
898
@section Memory model
899
@cindex memory model
900
 
901
The memory model used by the middle-end models that of the C/C++
902
languages.  The middle-end has the notion of an effective type
903
of a memory region which is used for type-based alias analysis.
904
 
905
The following is a refinement of ISO C99 6.5/6, clarifying the block copy case
906
to follow common sense and extending the concept of a dynamic effective
907
type to objects with a declared type as required for C++.
908
 
909
@smallexample
910
The effective type of an object for an access to its stored value is
911
the declared type of the object or the effective type determined by
912
a previous store to it.  If a value is stored into an object through
913
an lvalue having a type that is not a character type, then the
914
type of the lvalue becomes the effective type of the object for that
915
access and for subsequent accesses that do not modify the stored value.
916
If a value is copied into an object using @code{memcpy} or @code{memmove},
917
or is copied as an array of character type, then the effective type
918
of the modified object for that access and for subsequent accesses that
919
do not modify the value is undetermined.  For all other accesses to an
920
object, the effective type of the object is simply the type of the
921
lvalue used for the access.
922
@end smallexample
923
 

powered by: WebSVN 2.1.0

© copyright 1999-2024 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.