boehm-gc/doc/gcinterface.html

   1 <!DOCTYPE HTML>
   2 <HEAD>
   3 <TITLE>Garbage Collector Interface</TITLE>
   4 </HEAD>
   5 <BODY>
   6 <H1>C Interface</h1>
   7 On many platforms, a single-threaded garbage collector library can be built
   8 to act as a plug-in malloc replacement.  (Build with -DREDIRECT_MALLOC=GC_malloc
   9 -DIGNORE_FREE.)  This is often the best way to deal with third-party libraries
  10 which leak or prematurely free objects.  -DREDIRECT_MALLOC is intended
  11 primarily as an easy way to adapt old code, not for new development.
  12 <P>
  13 New code should use the interface discussed below.
  14 <P>
  15 Code must be linked against the GC library.  On most UNIX platforms,
  16 this will be gc.a.
  17 <P>
  18 The following describes the standard C interface to the garbage collector.
  19 It is not a complete definition of the interface.  It describes only the
  20 most commonly used functionality, approximately in decreasing order of
  21 frequency of use.  The description assumes an ANSI C compiler.
  22 The full interface is described in
  23 <A HREF="http://hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gch.txt">gc.h</a>
  24 or <TT>gc.h</tt> in the distribution.
  25 <P>
  26 Clients should include gc.h.
  27 <P>
  28 In the case of multithreaded code,
  29 gc.h should be included after the threads header file, and
  30 after defining the appropriate GC_XXXX_THREADS macro.
  31 (For 6.2alpha4 and later, simply defining GC_THREADS should suffice.)
  32 Gc.h must be included
  33 in files that use either GC or threads primitives, since threads primitives
  34 will be redefined to cooperate with the GC on many platforms.
  35 <DL>
  36 <DT> <B>void * GC_MALLOC(size_t <I>nbytes</i>)</b>
  37 <DD>
  38 Allocates and clears <I>nbytes</i> of storage.
  39 Requires (amortized) time proportional to <I>nbytes</i>.
  40 The resulting object will be automatically deallocated when unreferenced.
  41 References from objects allocated with the system malloc are usually not
  42 considered by the collector.  (See GC_MALLOC_UNCOLLECTABLE, however.)
  43 GC_MALLOC is a macro which invokes GC_malloc by default or, if GC_DEBUG
  44 is defined before gc.h is included, a debugging version that checks
  45 occasionally for overwrite errors, and the like.
  46 <DT> <B>void * GC_MALLOC_ATOMIC(size_t <I>nbytes</i>)</b>
  47 <DD>
  48 Allocates <I>nbytes</i> of storage.
  49 Requires (amortized) time proportional to <I>nbytes</i>.
  50 The resulting object will be automatically deallocated when unreferenced.
  51 The client promises that the resulting object will never contain any pointers.
  52 The memory is not cleared.
  53 This is the preferred way to allocate strings, floating point arrays,
  54 bitmaps, etc.
  55 More precise information about pointer locations can be communicated to the
  56 collector using the interface in
  57 <A HREF="http://www.hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gc_typedh.txt">gc_typed.h</a> in the distribution.
  58 <DT> <B>void * GC_MALLOC_UNCOLLECTABLE(size_t <I>nbytes</i>)</b>
  59 <DD>
  60 Identical to GC_MALLOC, except that the resulting object is not automatically
  61 deallocated.  Unlike the system-provided malloc, the collector does
  62 scan the object for pointers to garbage-collectable memory, even if the
  63 block itself does not appear to be reachable.  (Objects allocated in this way
  64 are effectively treated as roots by the collector.)
  65 <DT> <B> void * GC_REALLOC(void *old, size_t new_size) </b>
  66 <DD>
  67 Allocate a new object of the indicated size and copy (a prefix of) the
  68 old object into the new object.  The old object is reused in place if
  69 convenient.  If the original object was allocated with GC_malloc_atomic,
  70 the new object is subject to the same constraints.  If it was allocated
  71 as an uncollectable object, then the new object is uncollectable, and
  72 the old object (if different) is deallocated.
  73 (Use GC_REALLOC with GC_MALLOC, etc.)
  74 <DT> <B> void GC_FREE(void *dead) </b>
  75 <DD>
  76 Explicitly deallocate an object.  Typically not useful for small
  77 collectable objects.  (Use GC_FREE with GC_MALLOC, etc.)
  78 <DT> <B> void * GC_MALLOC_IGNORE_OFF_PAGE(size_t <I>nbytes</i>) </b>
  79 <DD>
  80 <DT> <B> void * GC_MALLOC_ATOMIC_IGNORE_OFF_PAGE(size_t <I>nbytes</i>) </b>
  81 <DD>
  82 Analogous to GC_MALLOC and GC_MALLOC_ATOMIC, except that the client
  83 guarantees that as long
  84 as the resulting object is of use, a pointer is maintained to someplace
  85 inside the first 512 bytes of the object.  This pointer should be declared
  86 volatile to avoid interference from compiler optimizations.
  87 (Other nonvolatile pointers to the object may exist as well.)
  88 This is the
  89 preferred way to allocate objects that are likely to be > 100KBytes in size.
  90 It greatly reduces the risk that such objects will be accidentally retained
  91 when they are no longer needed.  Thus space usage may be significantly reduced.
  92 <DT> <B> void GC_gcollect(void) </b>
  93 <DD>
  94 Explicitly force a garbage collection.
  95 <DT> <B> void GC_enable_incremental(void) </b>
  96 <DD>
  97 Cause the garbage collector to perform a small amount of work
  98 every few invocations of GC_malloc or the like, instead of performing
  99 an entire collection at once.  This is likely to increase total
 100 running time.  It will improve response on a platform that either has
 101 suitable support in the garbage collector (Irix and most other Unix
 102 versions, win32 if the collector was suitably built) or if "stubborn"
 103 allocation is used (see <A HREF="http://www.hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gch.txt">gc.h</a>).
 104 On many platforms this interacts poorly with system calls
 105 that write to the garbage collected heap.
 106 <DT> <B> GC_warn_proc GC_set_warn_proc(GC_warn_proc p) </b>
 107 <DD>
 108 Replace the default procedure used by the collector to print warnings.
 109 The collector
 110 may otherwise write to sterr, most commonly because GC_malloc was used
 111 in a situation in which GC_malloc_ignore_off_page would have been more
 112 appropriate.  See <A HREF="http://www.hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gch.txt">gc.h</a> for details.
 113 <DT> <B> void GC_register_finalizer(...) </b>
 114 <DD>
 115 Register a function to be called when an object becomes inaccessible.
 116 This is often useful as a backup method for releasing system resources
 117 (<I>e.g.</i> closing files) when the object referencing them becomes
 118 inaccessible.
 119 It is not an acceptable method to perform actions that must be performed
 120 in a timely fashion.
 121 See <A HREF="http://www.hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gch.txt">gc.h</a> for details of the interface.
 122 See <A HREF="http://www.hpl.hp.com/personal/Hans_Boehm/gc/finalization.html">here</a> for a more detailed discussion
 123 of the design.
 124 <P>
 125 Note that an object may become inaccessible before client code is done
 126 operating on its fields.  Suitable synchronization is usually required.
 127 See <A HREF="http://portal.acm.org/citation.cfm?doid=604131.604153">here</a>
 128 or <A HREF="http://www.hpl.hp.com/techreports/2002/HPL-2002-335.html">here</a>
 129 for details.
 130 </dl>
 131 <P>
 132 If you are concerned with multiprocessor performance and scalability,
 133 you should consider enabling and using thread local allocation (<I>e.g.</i>
 134 GC_LOCAL_MALLOC, see <TT>gc_local_alloc.h</tt>.  If your platform
 135 supports it, you should build the collector with parallel marking support
 136 (-DPARALLEL_MARK, or --enable-parallel-mark).
 137 <P>
 138 If the collector is used in an environment in which pointer location
 139 information for heap objects is easily available, this can be passed on
 140 to the colllector using the interfaces in either <TT>gc_typed.h</tt>
 141 or <TT>gc_gcj.h</tt>.
 142 <P>
 143 The collector distribution also includes a <B>string package</b> that takes
 144 advantage of the collector.  For details see
 145 <A HREF="http://www.hpl.hp.com/personal/Hans_Boehm/gc/gc_source/cordh.txt">cord.h</a>
 146
 147 <H1>C++ Interface</h1>
 148 There are three distinct ways to use the collector from C++:
 149 <DL>
 150 <DT> <B> STL allocators </b>
 151 <DD>
 152 Users of the <A HREF="http://www.sgi.com/tech/stl">SGI extended STL</a>
 153 can include <TT>new_gc_alloc.h</tt> before including
 154 STL header files.
 155 (<TT>gc_alloc.h</tt> corresponds to now obsolete versions of the
 156 SGI STL.)
 157 This defines SGI-style allocators
 158 <UL>
 159 <LI> alloc
 160 <LI> single_client_alloc
 161 <LI> gc_alloc
 162 <LI> single_client_gc_alloc
 163 </ul>
 164 which may be used either directly to allocate memory or to instantiate
 165 container templates.  The first two allocate uncollectable but traced
 166 memory, while the second two allocate collectable memory.
 167 The single_client versions are not safe for concurrent access by
 168 multiple threads, but are faster.
 169 <P>
 170 For an example, click <A HREF="http://hpl.hp.com/personal/Hans_Boehm/gc/gc_alloc_exC.txt">here</a>.
 171 <P>
 172 Recent versions of the collector also include a more standard-conforming
 173 allocator implemention in <TT>gc_allocator.h</tt>.  It defines
 174 <UL>
 175 <LI> traceable_allocator
 176 <LI> gc_allocator
 177 </ul>
 178 Again the former allocates uncollectable but traced memory.
 179 This should work with any fully standard-conforming C++ compiler.
 180 <DT> <B> Class inheritance based interface </b>
 181 <DD>
 182 Users may include gc_cpp.h and then cause members of certain classes to
 183 be allocated in garbage collectable memory by inheriting from class gc.
 184 For details see <A HREF="http://hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gc_cpph.txt">gc_cpp.h</a>.
 185 <DT> <B> C interface </b>
 186 <DD>
 187 It is also possible to use the C interface from
 188 <A HREF="http://hpl.hp.com/personal/Hans_Boehm/gc/gc_source/gch.txt">gc.h</a> directly.
 189 On platforms which use malloc to implement ::new, it should usually be possible
 190 to use a version of the collector that has been compiled as a malloc
 191 replacement.  It is also possible to replace ::new and other allocation
 192 functions suitably.
 193 <P>
 194 Note that user-implemented small-block allocation often works poorly with
 195 an underlying garbage-collected large block allocator, since the collector
 196 has to view all objects accessible from the user's free list as reachable.
 197 This is likely to cause problems if GC_malloc is used with something like
 198 the original HP version of STL.
 199 This approach works with the SGI versions of the STL only if the
 200 <TT>malloc_alloc</tt> allocator is used.
 201 </dl>
 202 </body>
 203 </html>