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

Subversion Repositories openrisc

[/] [openrisc/] [trunk/] [gnu-dev/] [or1k-gcc/] [gcc/] [doc/] [gty.texi] - Blame information for rev 747

Go to most recent revision | Details | Compare with Previous | View Log

Line No. Rev Author Line
1 711 jeremybenn
@c Copyright (C) 2002, 2003, 2004, 2007, 2008, 2009, 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
@node Type Information
7
@chapter Memory Management and Type Information
8
@cindex GGC
9
@findex GTY
10
 
11
GCC uses some fairly sophisticated memory management techniques, which
12
involve determining information about GCC's data structures from GCC's
13
source code and using this information to perform garbage collection and
14
implement precompiled headers.
15
 
16
A full C parser would be too complicated for this task, so a limited
17
subset of C is interpreted and special markers are used to determine
18
what parts of the source to look at.  All @code{struct} and
19
@code{union} declarations that define data structures that are
20
allocated under control of the garbage collector must be marked.  All
21
global variables that hold pointers to garbage-collected memory must
22
also be marked.  Finally, all global variables that need to be saved
23
and restored by a precompiled header must be marked.  (The precompiled
24
header mechanism can only save static variables if they're scalar.
25
Complex data structures must be allocated in garbage-collected memory
26
to be saved in a precompiled header.)
27
 
28
The full format of a marker is
29
@smallexample
30
GTY (([@var{option}] [(@var{param})], [@var{option}] [(@var{param})] @dots{}))
31
@end smallexample
32
@noindent
33
but in most cases no options are needed.  The outer double parentheses
34
are still necessary, though: @code{GTY(())}.  Markers can appear:
35
 
36
@itemize @bullet
37
@item
38
In a structure definition, before the open brace;
39
@item
40
In a global variable declaration, after the keyword @code{static} or
41
@code{extern}; and
42
@item
43
In a structure field definition, before the name of the field.
44
@end itemize
45
 
46
Here are some examples of marking simple data structures and globals.
47
 
48
@smallexample
49
struct GTY(()) @var{tag}
50
@{
51
  @var{fields}@dots{}
52
@};
53
 
54
typedef struct GTY(()) @var{tag}
55
@{
56
  @var{fields}@dots{}
57
@} *@var{typename};
58
 
59
static GTY(()) struct @var{tag} *@var{list};   /* @r{points to GC memory} */
60
static GTY(()) int @var{counter};        /* @r{save counter in a PCH} */
61
@end smallexample
62
 
63
The parser understands simple typedefs such as
64
@code{typedef struct @var{tag} *@var{name};} and
65
@code{typedef int @var{name};}.
66
These don't need to be marked.
67
 
68
@menu
69
* GTY Options::         What goes inside a @code{GTY(())}.
70
* GGC Roots::           Making global variables GGC roots.
71
* Files::               How the generated files work.
72
* Invoking the garbage collector::   How to invoke the garbage collector.
73
* Troubleshooting::     When something does not work as expected.
74
@end menu
75
 
76
@node GTY Options
77
@section The Inside of a @code{GTY(())}
78
 
79
Sometimes the C code is not enough to fully describe the type
80
structure.  Extra information can be provided with @code{GTY} options
81
and additional markers.  Some options take a parameter, which may be
82
either a string or a type name, depending on the parameter.  If an
83
option takes no parameter, it is acceptable either to omit the
84
parameter entirely, or to provide an empty string as a parameter.  For
85
example, @code{@w{GTY ((skip))}} and @code{@w{GTY ((skip ("")))}} are
86
equivalent.
87
 
88
When the parameter is a string, often it is a fragment of C code.  Four
89
special escapes may be used in these strings, to refer to pieces of
90
the data structure being marked:
91
 
92
@cindex % in GTY option
93
@table @code
94
@item %h
95
The current structure.
96
@item %1
97
The structure that immediately contains the current structure.
98
@item %0
99
The outermost structure that contains the current structure.
100
@item %a
101
A partial expression of the form @code{[i1][i2]@dots{}} that indexes
102
the array item currently being marked.
103
@end table
104
 
105
For instance, suppose that you have a structure of the form
106
@smallexample
107
struct A @{
108
  @dots{}
109
@};
110
struct B @{
111
  struct A foo[12];
112
@};
113
@end smallexample
114
@noindent
115
and @code{b} is a variable of type @code{struct B}.  When marking
116
@samp{b.foo[11]}, @code{%h} would expand to @samp{b.foo[11]},
117
@code{%0} and @code{%1} would both expand to @samp{b}, and @code{%a}
118
would expand to @samp{[11]}.
119
 
120
As in ordinary C, adjacent strings will be concatenated; this is
121
helpful when you have a complicated expression.
122
@smallexample
123
@group
124
GTY ((chain_next ("TREE_CODE (&%h.generic) == INTEGER_TYPE"
125
                  " ? TYPE_NEXT_VARIANT (&%h.generic)"
126
                  " : TREE_CHAIN (&%h.generic)")))
127
@end group
128
@end smallexample
129
 
130
The available options are:
131
 
132
@table @code
133
@findex length
134
@item length ("@var{expression}")
135
 
136
There are two places the type machinery will need to be explicitly told
137
the length of an array.  The first case is when a structure ends in a
138
variable-length array, like this:
139
@smallexample
140
struct GTY(()) rtvec_def @{
141
  int num_elem;         /* @r{number of elements} */
142
  rtx GTY ((length ("%h.num_elem"))) elem[1];
143
@};
144
@end smallexample
145
 
146
In this case, the @code{length} option is used to override the specified
147
array length (which should usually be @code{1}).  The parameter of the
148
option is a fragment of C code that calculates the length.
149
 
150
The second case is when a structure or a global variable contains a
151
pointer to an array, like this:
152
@smallexample
153
struct gimple_omp_for_iter * GTY((length ("%h.collapse"))) iter;
154
@end smallexample
155
In this case, @code{iter} has been allocated by writing something like
156
@smallexample
157
  x->iter = ggc_alloc_cleared_vec_gimple_omp_for_iter (collapse);
158
@end smallexample
159
and the @code{collapse} provides the length of the field.
160
 
161
This second use of @code{length} also works on global variables, like:
162
@verbatim
163
static GTY((length("reg_known_value_size"))) rtx *reg_known_value;
164
@end verbatim
165
 
166
@findex skip
167
@item skip
168
 
169
If @code{skip} is applied to a field, the type machinery will ignore it.
170
This is somewhat dangerous; the only safe use is in a union when one
171
field really isn't ever used.
172
 
173
@findex desc
174
@findex tag
175
@findex default
176
@item desc ("@var{expression}")
177
@itemx tag ("@var{constant}")
178
@itemx default
179
 
180
The type machinery needs to be told which field of a @code{union} is
181
currently active.  This is done by giving each field a constant
182
@code{tag} value, and then specifying a discriminator using @code{desc}.
183
The value of the expression given by @code{desc} is compared against
184
each @code{tag} value, each of which should be different.  If no
185
@code{tag} is matched, the field marked with @code{default} is used if
186
there is one, otherwise no field in the union will be marked.
187
 
188
In the @code{desc} option, the ``current structure'' is the union that
189
it discriminates.  Use @code{%1} to mean the structure containing it.
190
There are no escapes available to the @code{tag} option, since it is a
191
constant.
192
 
193
For example,
194
@smallexample
195
struct GTY(()) tree_binding
196
@{
197
  struct tree_common common;
198
  union tree_binding_u @{
199
    tree GTY ((tag ("0"))) scope;
200
    struct cp_binding_level * GTY ((tag ("1"))) level;
201
  @} GTY ((desc ("BINDING_HAS_LEVEL_P ((tree)&%0)"))) xscope;
202
  tree value;
203
@};
204
@end smallexample
205
 
206
In this example, the value of BINDING_HAS_LEVEL_P when applied to a
207
@code{struct tree_binding *} is presumed to be 0 or 1.  If 1, the type
208
mechanism will treat the field @code{level} as being present and if 0,
209
will treat the field @code{scope} as being present.
210
 
211
@findex param_is
212
@findex use_param
213
@item param_is (@var{type})
214
@itemx use_param
215
 
216
Sometimes it's convenient to define some data structure to work on
217
generic pointers (that is, @code{PTR}) and then use it with a specific
218
type.  @code{param_is} specifies the real type pointed to, and
219
@code{use_param} says where in the generic data structure that type
220
should be put.
221
 
222
For instance, to have a @code{htab_t} that points to trees, one would
223
write the definition of @code{htab_t} like this:
224
@smallexample
225
typedef struct GTY(()) @{
226
  @dots{}
227
  void ** GTY ((use_param, @dots{})) entries;
228
  @dots{}
229
@} htab_t;
230
@end smallexample
231
and then declare variables like this:
232
@smallexample
233
  static htab_t GTY ((param_is (union tree_node))) ict;
234
@end smallexample
235
 
236
@findex param@var{n}_is
237
@findex use_param@var{n}
238
@item param@var{n}_is (@var{type})
239
@itemx use_param@var{n}
240
 
241
In more complicated cases, the data structure might need to work on
242
several different types, which might not necessarily all be pointers.
243
For this, @code{param1_is} through @code{param9_is} may be used to
244
specify the real type of a field identified by @code{use_param1} through
245
@code{use_param9}.
246
 
247
@findex use_params
248
@item use_params
249
 
250
When a structure contains another structure that is parameterized,
251
there's no need to do anything special, the inner structure inherits the
252
parameters of the outer one.  When a structure contains a pointer to a
253
parameterized structure, the type machinery won't automatically detect
254
this (it could, it just doesn't yet), so it's necessary to tell it that
255
the pointed-to structure should use the same parameters as the outer
256
structure.  This is done by marking the pointer with the
257
@code{use_params} option.
258
 
259
@findex deletable
260
@item deletable
261
 
262
@code{deletable}, when applied to a global variable, indicates that when
263
garbage collection runs, there's no need to mark anything pointed to
264
by this variable, it can just be set to @code{NULL} instead.  This is used
265
to keep a list of free structures around for re-use.
266
 
267
@findex if_marked
268
@item if_marked ("@var{expression}")
269
 
270
Suppose you want some kinds of object to be unique, and so you put them
271
in a hash table.  If garbage collection marks the hash table, these
272
objects will never be freed, even if the last other reference to them
273
goes away.  GGC has special handling to deal with this: if you use the
274
@code{if_marked} option on a global hash table, GGC will call the
275
routine whose name is the parameter to the option on each hash table
276
entry.  If the routine returns nonzero, the hash table entry will
277
be marked as usual.  If the routine returns zero, the hash table entry
278
will be deleted.
279
 
280
The routine @code{ggc_marked_p} can be used to determine if an element
281
has been marked already; in fact, the usual case is to use
282
@code{if_marked ("ggc_marked_p")}.
283
 
284
@findex mark_hook
285
@item mark_hook ("@var{hook-routine-name}")
286
 
287
If provided for a structure or union type, the given
288
@var{hook-routine-name} (between double-quotes) is the name of a
289
routine called when the garbage collector has just marked the data as
290
reachable. This routine should not change the data, or call any ggc
291
routine. Its only argument is a pointer to the just marked (const)
292
structure or union.
293
 
294
@findex maybe_undef
295
@item maybe_undef
296
 
297
When applied to a field, @code{maybe_undef} indicates that it's OK if
298
the structure that this fields points to is never defined, so long as
299
this field is always @code{NULL}.  This is used to avoid requiring
300
backends to define certain optional structures.  It doesn't work with
301
language frontends.
302
 
303
@findex nested_ptr
304
@item nested_ptr (@var{type}, "@var{to expression}", "@var{from expression}")
305
 
306
The type machinery expects all pointers to point to the start of an
307
object.  Sometimes for abstraction purposes it's convenient to have
308
a pointer which points inside an object.  So long as it's possible to
309
convert the original object to and from the pointer, such pointers
310
can still be used.  @var{type} is the type of the original object,
311
the @var{to expression} returns the pointer given the original object,
312
and the @var{from expression} returns the original object given
313
the pointer.  The pointer will be available using the @code{%h}
314
escape.
315
 
316
@findex chain_next
317
@findex chain_prev
318
@findex chain_circular
319
@item chain_next ("@var{expression}")
320
@itemx chain_prev ("@var{expression}")
321
@itemx chain_circular ("@var{expression}")
322
 
323
It's helpful for the type machinery to know if objects are often
324
chained together in long lists; this lets it generate code that uses
325
less stack space by iterating along the list instead of recursing down
326
it.  @code{chain_next} is an expression for the next item in the list,
327
@code{chain_prev} is an expression for the previous item.  For singly
328
linked lists, use only @code{chain_next}; for doubly linked lists, use
329
both.  The machinery requires that taking the next item of the
330
previous item gives the original item.  @code{chain_circular} is similar
331
to @code{chain_next}, but can be used for circular single linked lists.
332
 
333
@findex reorder
334
@item reorder ("@var{function name}")
335
 
336
Some data structures depend on the relative ordering of pointers.  If
337
the precompiled header machinery needs to change that ordering, it
338
will call the function referenced by the @code{reorder} option, before
339
changing the pointers in the object that's pointed to by the field the
340
option applies to.  The function must take four arguments, with the
341
signature @samp{@w{void *, void *, gt_pointer_operator, void *}}.
342
The first parameter is a pointer to the structure that contains the
343
object being updated, or the object itself if there is no containing
344
structure.  The second parameter is a cookie that should be ignored.
345
The third parameter is a routine that, given a pointer, will update it
346
to its correct new value.  The fourth parameter is a cookie that must
347
be passed to the second parameter.
348
 
349
PCH cannot handle data structures that depend on the absolute values
350
of pointers.  @code{reorder} functions can be expensive.  When
351
possible, it is better to depend on properties of the data, like an ID
352
number or the hash of a string instead.
353
 
354
@findex variable_size
355
@item variable_size
356
 
357
The type machinery expects the types to be of constant size.  When this
358
is not true, for example, with structs that have array fields or unions,
359
the type machinery cannot tell how many bytes need to be allocated at
360
each allocation.  The @code{variable_size} is used to mark such types.
361
The type machinery then provides allocators that take a parameter
362
indicating an exact size of object being allocated.  Note that the size
363
must be provided in bytes whereas the @code{length} option works with
364
array lengths in number of elements.
365
 
366
For example,
367
@smallexample
368
struct GTY((variable_size)) sorted_fields_type @{
369
  int len;
370
  tree GTY((length ("%h.len"))) elts[1];
371
@};
372
@end smallexample
373
 
374
Then the objects of @code{struct sorted_fields_type} are allocated in GC
375
memory as follows:
376
@smallexample
377
  field_vec = ggc_alloc_sorted_fields_type (size);
378
@end smallexample
379
 
380
If @var{field_vec->elts} stores @var{n} elements, then @var{size}
381
could be calculated as follows:
382
@smallexample
383
  size_t size = sizeof (struct sorted_fields_type) + n * sizeof (tree);
384
@end smallexample
385
 
386
@findex atomic
387
@item atomic
388
 
389
The @code{atomic} option can only be used with pointers.  It informs
390
the GC machinery that the memory that the pointer points to does not
391
contain any pointers, and hence it should be treated by the GC and PCH
392
machinery as an ``atomic'' block of memory that does not need to be
393
examined when scanning memory for pointers.  In particular, the
394
machinery will not scan that memory for pointers to mark them as
395
reachable (when marking pointers for GC) or to relocate them (when
396
writing a PCH file).
397
 
398
The @code{atomic} option differs from the @code{skip} option.
399
@code{atomic} keeps the memory under Garbage Collection, but makes the
400
GC ignore the contents of the memory.  @code{skip} is more drastic in
401
that it causes the pointer and the memory to be completely ignored by
402
the Garbage Collector.  So, memory marked as @code{atomic} is
403
automatically freed when no longer reachable, while memory marked as
404
@code{skip} is not.
405
 
406
The @code{atomic} option must be used with great care, because all
407
sorts of problem can occur if used incorrectly, that is, if the memory
408
the pointer points to does actually contain a pointer.
409
 
410
Here is an example of how to use it:
411
@smallexample
412
struct GTY(()) my_struct @{
413
  int number_of_elements;
414
  unsigned int GTY ((atomic)) * elements;
415
@};
416
@end smallexample
417
In this case, @code{elements} is a pointer under GC, and the memory it
418
points to needs to be allocated using the Garbage Collector, and will
419
be freed automatically by the Garbage Collector when it is no longer
420
referenced.  But the memory that the pointer points to is an array of
421
@code{unsigned int} elements, and the GC must not try to scan it to
422
find pointers to mark or relocate, which is why it is marked with the
423
@code{atomic} option.
424
 
425
Note that, currently, global variables can not be marked with
426
@code{atomic}; only fields of a struct can.  This is a known
427
limitation.  It would be useful to be able to mark global pointers
428
with @code{atomic} to make the PCH machinery aware of them so that
429
they are saved and restored correctly to PCH files.
430
 
431
@findex special
432
@item special ("@var{name}")
433
 
434
The @code{special} option is used to mark types that have to be dealt
435
with by special case machinery.  The parameter is the name of the
436
special case.  See @file{gengtype.c} for further details.  Avoid
437
adding new special cases unless there is no other alternative.
438
@end table
439
 
440
@node GGC Roots
441
@section Marking Roots for the Garbage Collector
442
@cindex roots, marking
443
@cindex marking roots
444
 
445
In addition to keeping track of types, the type machinery also locates
446
the global variables (@dfn{roots}) that the garbage collector starts
447
at.  Roots must be declared using one of the following syntaxes:
448
 
449
@itemize @bullet
450
@item
451
@code{extern GTY(([@var{options}])) @var{type} @var{name};}
452
@item
453
@code{static GTY(([@var{options}])) @var{type} @var{name};}
454
@end itemize
455
@noindent
456
The syntax
457
@itemize @bullet
458
@item
459
@code{GTY(([@var{options}])) @var{type} @var{name};}
460
@end itemize
461
@noindent
462
is @emph{not} accepted.  There should be an @code{extern} declaration
463
of such a variable in a header somewhere---mark that, not the
464
definition.  Or, if the variable is only used in one file, make it
465
@code{static}.
466
 
467
@node Files
468
@section Source Files Containing Type Information
469
@cindex generated files
470
@cindex files, generated
471
 
472
Whenever you add @code{GTY} markers to a source file that previously
473
had none, or create a new source file containing @code{GTY} markers,
474
there are three things you need to do:
475
 
476
@enumerate
477
@item
478
You need to add the file to the list of source files the type
479
machinery scans.  There are four cases:
480
 
481
@enumerate a
482
@item
483
For a back-end file, this is usually done
484
automatically; if not, you should add it to @code{target_gtfiles} in
485
the appropriate port's entries in @file{config.gcc}.
486
 
487
@item
488
For files shared by all front ends, add the filename to the
489
@code{GTFILES} variable in @file{Makefile.in}.
490
 
491
@item
492
For files that are part of one front end, add the filename to the
493
@code{gtfiles} variable defined in the appropriate
494
@file{config-lang.in}.  For C, the file is @file{c-config-lang.in}.
495
Headers should appear before non-headers in this list.
496
 
497
@item
498
For files that are part of some but not all front ends, add the
499
filename to the @code{gtfiles} variable of @emph{all} the front ends
500
that use it.
501
@end enumerate
502
 
503
@item
504
If the file was a header file, you'll need to check that it's included
505
in the right place to be visible to the generated files.  For a back-end
506
header file, this should be done automatically.  For a front-end header
507
file, it needs to be included by the same file that includes
508
@file{gtype-@var{lang}.h}.  For other header files, it needs to be
509
included in @file{gtype-desc.c}, which is a generated file, so add it to
510
@code{ifiles} in @code{open_base_file} in @file{gengtype.c}.
511
 
512
For source files that aren't header files, the machinery will generate a
513
header file that should be included in the source file you just changed.
514
The file will be called @file{gt-@var{path}.h} where @var{path} is the
515
pathname relative to the @file{gcc} directory with slashes replaced by
516
@verb{|-|}, so for example the header file to be included in
517
@file{cp/parser.c} is called @file{gt-cp-parser.c}.  The
518
generated header file should be included after everything else in the
519
source file.  Don't forget to mention this file as a dependency in the
520
@file{Makefile}!
521
 
522
@end enumerate
523
 
524
For language frontends, there is another file that needs to be included
525
somewhere.  It will be called @file{gtype-@var{lang}.h}, where
526
@var{lang} is the name of the subdirectory the language is contained in.
527
 
528
Plugins can add additional root tables.  Run the @code{gengtype}
529
utility in plugin mode as @code{gengtype -P pluginout.h @var{source-dir}
530
@var{file-list} @var{plugin*.c}} with your plugin files
531
@var{plugin*.c} using @code{GTY} to generate the @var{pluginout.h} file.
532
The GCC build tree is needed to be present in that mode.
533
 
534
 
535
@node Invoking the garbage collector
536
@section How to invoke the garbage collector
537
@cindex garbage collector, invocation
538
@findex ggc_collect
539
 
540
The GCC garbage collector GGC is only invoked explicitly. In contrast
541
with many other garbage collectors, it is not implicitly invoked by
542
allocation routines when a lot of memory has been consumed. So the
543
only way to have GGC reclaim storage it to call the @code{ggc_collect}
544
function explicitly.  This call is an expensive operation, as it may
545
have to scan the entire heap.  Beware that local variables (on the GCC
546
call stack) are not followed by such an invocation (as many other
547
garbage collectors do): you should reference all your data from static
548
or external @code{GTY}-ed variables, and it is advised to call
549
@code{ggc_collect} with a shallow call stack.  The GGC is an exact mark
550
and sweep garbage collector (so it does not scan the call stack for
551
pointers).  In practice GCC passes don't often call @code{ggc_collect}
552
themselves, because it is called by the pass manager between passes.
553
 
554
At the time of the @code{ggc_collect} call all pointers in the GC-marked
555
structures must be valid or @code{NULL}.  In practice this means that
556
there should not be uninitialized pointer fields in the structures even
557
if your code never reads or writes those fields at a particular
558
instance.  One way to ensure this is to use cleared versions of
559
allocators unless all the fields are initialized manually immediately
560
after allocation.
561
 
562
@node Troubleshooting
563
@section Troubleshooting the garbage collector
564
@cindex garbage collector, troubleshooting
565
 
566
With the current garbage collector implementation, most issues should
567
show up as GCC compilation errors.  Some of the most commonly
568
encountered issues are described below.
569
 
570
@itemize @bullet
571
@item Gengtype does not produce allocators for a @code{GTY}-marked type.
572
Gengtype checks if there is at least one possible path from GC roots to
573
at least one instance of each type before outputting allocators.  If
574
there is no such path, the @code{GTY} markers will be ignored and no
575
allocators will be output.  Solve this by making sure that there exists
576
at least one such path.  If creating it is unfeasible or raises a ``code
577
smell'', consider if you really must use GC for allocating such type.
578
 
579
@item Link-time errors about undefined @code{gt_ggc_r_foo_bar} and
580
similarly-named symbols.  Check if your @file{foo_bar} source file has
581
@code{#include "gt-foo_bar.h"} as its very last line.
582
 
583
@end itemize

powered by: WebSVN 2.1.0

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