1 The collector uses a large amount of conditional compilation in order to
2 deal with platform dependencies. This violates a number of known coding
3 standards. On the other hand, it seems to be the only practical way to
4 support this many platforms without excessive code duplication.
6 A few guidelines have mostly been followed in order to keep this manageable:
8 1) #if and #ifdef directives are properly indented whenever easily possible.
9 All known C compilers allow whitespace between the "#" and the "if" to make
10 this possible. ANSI C also allows white space before the "#", though we
11 avoid that. It has the known disadvantages that it differs from the normal
12 GNU conventions, and that it makes patches larger than otherwise necessary.
13 In my opinion, it's still well worth it, for the same reason that we indent
14 ordinary "if" statements.
16 2) Whenever possible, tests are performed on the macros defined in gcconfig.h
17 instead of directly testing platform-specific predefined macros. This makes
18 it relatively easy to adapt to new compilers with a different set of
19 predefined macros. Currently these macros generally identify platforms
20 instead of features. In many cases, this is a mistake.
22 Many of the tested configuration macros are at least somewhat defined in
23 either include/private/gcconfig.h or in Makefile.direct. Here is an attempt
24 at documenting these macros: (Thanks to Walter Bright for suggesting
25 this. This is a work in progress)
30 GC_DEBUG Tested by gc.h. Causes all-upper-case macros to
31 expand to calls to debug versions of collector routines.
33 GC_DEBUG_REPLACEMENT Tested by gc.h. Causes GC_MALLOC/REALLOC() to be
34 defined as GC_debug_malloc/realloc_replacement().
36 GC_NO_THREAD_REDIRECTS Tested by gc.h. Prevents redirection of thread
37 creation routines etc. to GC_ versions. Requires the
38 programmer to explicitly handle thread registration.
40 GC_NO_THREAD_DECLS Tested by gc.h. MS Windows only. Do not declare
41 Windows thread creation routines and do not include windows.h.
43 GC_UNDERSCORE_STDCALL Tested by gc.h. Explicitly prefix exported/imported
44 WINAPI (__stdcall) symbols with '_' (underscore). Could be
45 used with MinGW (for x86) compiler (in conjunction with
46 GC_DLL) to follow MS conventions for __stdcall symbols naming.
49 #define'd by the Digital Mars C++ compiler when
50 operator new[] and delete[] are separately
51 overloadable. Used in gc_cpp.h.
53 _DLL Tested by gc_config_macros.h. Defined by Visual C++ if runtime
54 dynamic libraries are in use. Used (only if none of GC_DLL,
55 GC_NOT_DLL, __GNUC__ are defined) to test whether
56 __declspec(dllimport) needs to be added to declarations
57 to support the case in which the collector is in a DLL.
59 GC_DLL Defined by user if dynamic libraries are being built
60 or used. Also set by gc.h if _DLL is defined (except for
61 mingw) while GC_NOT_DLL and __GNUC__ are both undefined.
62 This is the macro that is tested internally to determine
63 whether the GC is in its own dynamic library. May need
64 to be set by clients before including gc.h. Note that
65 inside the GC implementation it indicates that the
66 collector is in its own dynamic library, should export
67 its symbols, etc. But in clients it indicates that the
68 GC resides in a different DLL, its entry points should
69 be referenced accordingly, and precautions may need to
70 be taken to properly deal with statically allocated
71 variables in the main program. Used for MS Windows.
72 Also used by GCC v4+ (only when the dynamic shared library
73 is being built) to hide internally used symbols.
75 GC_NOT_DLL User-settable macro that overrides _DLL, e.g. if runtime
76 dynamic libraries are used, but the collector is in a static
77 library. Tested by gc_config_macros.h.
79 GC_REQUIRE_WCSDUP Force GC to export GC_wcsdup() (the Unicode version
80 of GC_strdup); could be useful in the leak-finding mode.
83 These define arguments influence the collector configuration:
85 FIND_LEAK Causes GC_find_leak to be initially set. This causes the
86 collector to assume that all inaccessible objects should have been
87 explicitly deallocated, and reports exceptions. Finalization and the test
88 program are not usable in this mode.
90 GC_FINDLEAK_DELAY_FREE Turns on deferred freeing of objects in the
91 leak-finding mode letting the collector to detect alter-object-after-free
92 errors as well as detect leaked objects sooner (instead of only when program
93 terminates). Has no effect if SHORT_DBG_HDRS.
95 GC_ABORT_ON_LEAK Causes the application to be terminated once leaked or
96 smashed (corrupted on use-after-free) objects are found (after printing the
97 information about that objects).
99 SUNOS5SIGS Solaris-like signal handling. This is probably misnamed,
100 since it really doesn't guarantee much more than POSIX. Currently set only
101 for Solaris2.X, HPUX, and DRSNX. Should probably be set for some other
104 PCR Set if the collector is being built as part of the Xerox Portable
107 USE_COMPILER_TLS Assume the existence of __thread-style thread-local storage.
108 Set automatically for thread-local allocation with the HP/UX vendor
109 compiler. Usable with gcc on sufficiently up-to-date ELF platforms.
111 IMPORTANT: Any of the _THREADS options must normally also be defined in
112 the client before including gc.h. This redefines thread primitives to
113 invoke the GC_ versions instead. Alternatively, linker-based symbol
114 interception can be used on a few platforms.
116 GC_THREADS Should set the appropriate one of the below macros,
117 except GC_WIN32_PTHREADS, which must be set explicitly. Tested by gc.h.
119 GC_SOLARIS_THREADS Enables support for Solaris pthreads.
120 Must also define _REENTRANT.
122 GC_IRIX_THREADS Enables support for Irix pthreads. See README.sgi.
124 GC_HPUX_THREADS Enables support for HP/UX 11 pthreads.
125 Also requires _REENTRANT or _POSIX_C_SOURCE=199506L. See README.hp.
127 GC_LINUX_THREADS Enables support for Xavier Leroy's Linux threads
128 or NPTL threads. See README.linux. _REENTRANT may also be required.
130 GC_OSF1_THREADS Enables support for Tru64 pthreads.
132 GC_FREEBSD_THREADS Enables support for FreeBSD pthreads.
133 Appeared to run into some underlying thread problems.
135 GC_NETBSD_THREADS Enables support for NetBSD pthreads.
137 GC_OPENBSD_THREADS Enables support for OpenBSD pthreads.
139 GC_DARWIN_THREADS Enables support for Mac OS X pthreads.
141 GC_AIX_THREADS Enables support for IBM AIX threads.
143 GC_DGUX386_THREADS Enables support for DB/UX on I386 threads.
144 See README.DGUX386. (Probably has not been tested recently.)
146 GC_WIN32_THREADS Enables support for Win32 threads. That makes sense
147 for this Makefile only under Cygwin.
149 GC_WIN32_PTHREADS Enables support for Ming32 pthreads. This cannot be
150 enabled automatically by GC_THREADS, which would assume Win32 native
153 PTW32_STATIC_LIB Causes the static version of the Mingw pthreads
154 library to be used. Requires GC_WIN32_PTHREADS.
156 GC_PTHREADS_PARAMARK Causes pthread-based parallel mark implementation
157 to be used even if GC_WIN32_PTHREADS is undefined. (Useful for WinCE.)
159 ALL_INTERIOR_POINTERS Allows all pointers to the interior of objects to be
160 recognized. (See gc_priv.h for consequences.) Alternatively,
161 GC_all_interior_pointers can be set at process initialization time.
163 SMALL_CONFIG Tries to tune the collector for small heap sizes,
164 usually causing it to use less space in such situations. Incremental
165 collection no longer works in this case. Also, removes some
166 statistic-printing code. Turns off some optimization algorithms (like data
167 prefetching in the mark routine).
169 GC_DISABLE_INCREMENTAL Turn off the incremental collection support.
171 NO_INCREMENTAL Causes the gctest program to not invoke the incremental
172 collector. This has no impact on the generated library, only on the test
173 program. (This is often useful for debugging failures unrelated to
176 LARGE_CONFIG Tunes the collector for unusually large heaps.
177 Necessary for heaps larger than about 4 GiB on most (64-bit) machines.
178 Recommended for heaps larger than about 500 MiB. Not recommended for
179 embedded systems. Could be used in conjunction with SMALL_CONFIG to
180 generate smaller code (by disabling incremental collection support,
181 statistic printing and some optimization algorithms).
183 DONT_ADD_BYTE_AT_END Meaningful only with ALL_INTERIOR_POINTERS or
184 GC_all_interior_pointers = 1. Normally ALL_INTERIOR_POINTERS
185 causes all objects to be padded so that pointers just past the end of
186 an object can be recognized. This can be expensive. (The padding
187 is normally more than one byte due to alignment constraints.)
188 DONT_ADD_BYTE_AT_END disables the padding.
190 NO_EXECUTE_PERMISSION May cause some or all of the heap to not
191 have execute permission, i.e. it may be impossible to execute
192 code from the heap. Currently this only affects the incremental
193 collector on UNIX machines. It may greatly improve its performance,
194 since this may avoid some expensive cache synchronization. Alternatively,
195 GC_set_pages_executable can be called at the process initialization time.
197 GC_NO_OPERATOR_NEW_ARRAY Declares that the C++ compiler does not
198 support the new syntax "operator new[]" for allocating and deleting arrays.
199 See gc_cpp.h for details. No effect on the C part of the collector.
200 This is defined implicitly in a few environments. Must also be defined
201 by clients that use gc_cpp.h.
203 REDIRECT_MALLOC=<X> Causes malloc to be defined as alias for X.
204 Unless the following macros are defined, realloc is also redirected
205 to GC_realloc, and free is redirected to GC_free.
206 Calloc and str[n]dup are redefined in terms of the new malloc. X should
207 be either GC_malloc or GC_malloc_uncollectable, or
208 GC_debug_malloc_replacement. (The latter invokes GC_debug_malloc
209 with dummy source location information, but still results in
210 properly remembered call stacks on Linux/X86 and Solaris/SPARC.
211 It requires that the following two macros also be used.)
212 The former is occasionally useful for working around leaks in code
213 you don't want to (or can't) look at. It may not work for
214 existing code, but it often does. Neither works on all platforms,
215 since some ports use malloc or calloc to obtain system memory.
216 (Probably works for UNIX, and Win32.) If you build with DBG_HDRS_ALL,
217 you should only use GC_debug_malloc_replacement as a malloc
220 REDIRECT_REALLOC=<X> Causes GC_realloc to be redirected to X.
221 The canonical use is REDIRECT_REALLOC=GC_debug_realloc_replacement,
222 together with REDIRECT_MALLOC=GC_debug_malloc_replacement to
223 generate leak reports with call stacks for both malloc and realloc.
224 This also requires REDIRECT_FREE.
226 REDIRECT_FREE=<X> Causes free to be redirected to X. The canonical use
227 is REDIRECT_FREE=GC_debug_free.
229 IGNORE_FREE Turns calls to free into a no-op. Only useful with
232 NO_DEBUGGING Removes GC_dump and the debugging routines it calls.
233 Reduces code size slightly at the expense of debuggability.
235 DEBUG_THREADS Turn on printing additional thread-support debugging
238 JAVA_FINALIZATION Makes it somewhat safer to finalize objects out of
239 order by specifying a nonstandard finalization mark procedure (see
240 finalize.c). Objects reachable from finalizable objects will be marked
241 in a separate post-pass, and hence their memory won't be reclaimed.
242 Not recommended unless you are implementing a language that specifies
243 these semantics. Since 5.0, determines only the initial value
244 of GC_java_finalization variable.
246 FINALIZE_ON_DEMAND Causes finalizers to be run only in response
247 to explicit GC_invoke_finalizers() calls.
248 In 5.0 this became runtime adjustable, and this only determines the
249 initial value of GC_finalize_on_demand.
251 ATOMIC_UNCOLLECTABLE Includes code for GC_malloc_atomic_uncollectable.
252 This is useful if either the vendor malloc implementation is poor,
253 or if REDIRECT_MALLOC is used.
255 MARK_BIT_PER_GRANULE Requests that a mark bit (or often byte)
256 be allocated for each allocation granule, as opposed to each object.
257 This often improves speed, possibly at some cost in space and/or
258 cache footprint. Normally it is best to let this decision be
259 made automatically depending on platform.
261 MARK_BIT_PER_OBJ Requests that a mark bit be allocated for each
262 object instead of allocation granule. The opposite of
263 MARK_BIT_PER_GRANULE.
265 HBLKSIZE=<ddd> Explicitly sets the heap block size (where ddd is a power of
266 2 between 512 and 16384). Each heap block is devoted to a single size and
267 kind of object. For the incremental collector it makes sense to match
268 the most likely page size. Otherwise large values result in more
269 fragmentation, but generally better performance for large heaps.
271 USE_MMAP Use MMAP instead of sbrk to get new memory.
272 Works for Linux, FreeBSD, Cygwin, Solaris and Irix.
274 USE_MUNMAP Causes memory to be returned to the OS under the right
275 circumstances. This currently disables VM-based incremental collection
276 (except for Win32 with GetWriteWatch() available).
277 Works under some Unix, Linux and Windows versions.
278 Requires USE_MMAP except for Windows.
280 MUNMAP_THRESHOLD=<value> Set the desired memory blocks unmapping
281 threshold (the number of sequential garbage collections for which
282 a candidate block for unmapping should remain free).
284 GC_FORCE_UNMAP_ON_GCOLLECT Set "unmap as much as possible on explicit GC"
285 mode on by default. The mode could be changed at run-time. Has no effect
286 unless unmapping is turned on. Has no effect on implicitly-initiated
289 MMAP_STACKS (for Solaris threads) Use mmap from /dev/zero rather than
290 GC_scratch_alloc() to get stack memory.
292 PRINT_BLACK_LIST Whenever a black list entry is added, i.e. whenever
293 the garbage collector detects a value that looks almost, but not quite,
294 like a pointer, print both the address containing the value, and the
295 value of the near-bogus-pointer. Can be used to identify regions of
296 memory that are likely to contribute misidentified pointers.
298 KEEP_BACK_PTRS Add code to save back pointers in debugging headers
299 for objects allocated with the debugging allocator. If all objects
300 through GC_MALLOC with GC_DEBUG defined, this allows the client
301 to determine how particular or randomly chosen objects are reachable
302 for debugging/profiling purposes. The gc_backptr.h interface is
303 implemented only if this is defined.
305 GC_ASSERTIONS Enable some internal GC assertion checking. Currently
306 this facility is only used in a few places. It is intended primarily
307 for debugging of the garbage collector itself, but could also...
309 DBG_HDRS_ALL Make sure that all objects have debug headers. Increases
310 the reliability (from 99.9999% to 100% mod. bugs) of some of the debugging
311 code (especially KEEP_BACK_PTRS). Makes SHORT_DBG_HDRS possible.
312 Assumes that all client allocation is done through debugging allocators.
314 SHORT_DBG_HDRS Assume that all objects have debug headers. Shorten
315 the headers to minimize object size, at the expense of checking for
316 writes past the end of an object. This is intended for environments
317 in which most client code is written in a "safe" language, such as
318 Scheme or Java. Assumes that all client allocation is done using
319 the GC_debug_ functions, or through the macros that expand to these,
320 or by redirecting malloc to GC_debug_malloc_replacement.
321 (Also eliminates the field for the requested object size.)
322 Occasionally could be useful for debugging of client code. Slows down the
323 collector somewhat, but not drastically.
325 SAVE_CALL_COUNT=<n> Set the number of call frames saved with objects
326 allocated through the debugging interface. Affects the amount of
327 information generated in leak reports. Only matters on platforms
328 on which we can quickly generate call stacks, currently Linux/(X86 & SPARC)
329 and Solaris/SPARC and platforms that provide execinfo.h.
330 Default is zero. On X86, client
331 code should NOT be compiled with -fomit-frame-pointer.
333 SAVE_CALL_NARGS=<n> Set the number of functions arguments to be saved
334 with each call frame. Default is zero. Ignored if we don't know how to
335 retrieve arguments on the platform.
337 CHECKSUMS Reports on erroneously clear dirty bits, and unexpectedly
338 altered stubborn objects, at substantial performance cost. Use only for
339 debugging of the incremental collector. Not compatible with USE_MUNMAP
342 GC_GCJ_SUPPORT Includes support for gcj (and possibly other systems
343 that include a pointer to a type descriptor in each allocated object).
344 Building this way requires an ANSI C compiler.
346 USE_I686_PREFETCH Causes the collector to issue Pentium III style
347 prefetch instructions. No effect except on X86 Linux platforms.
348 Assumes a very recent gcc-compatible compiler and assembler.
349 (Gas prefetcht0 support was added around May 1999.)
350 Empirically the code appears to still run correctly on Pentium II
351 processors, though with no performance benefit. May not run on other
352 X86 processors? In some cases this improves performance by
355 USE_3DNOW_PREFETCH Causes the collector to issue AMD 3DNow style
356 prefetch instructions. Same restrictions as USE_I686_PREFETCH.
357 Minimally tested. Didn't appear to be an obvious win on a K6-2/500.
359 USE_PPC_PREFETCH Causes the collector to issue PowerPC style
360 prefetch instructions. No effect except on PowerPC OS X platforms.
361 Performance impact untested.
363 GC_USE_LD_WRAP In combination with the old flags listed in README.linux
364 causes the collector some system and pthread calls in a more transparent
365 fashion than the usual macro-based approach. Requires GNU ld, and
366 currently probably works only with Linux.
368 GC_USE_DLOPEN_WRAP Causes the collector to redefine malloc and
369 intercepted pthread routines with their real names, and causes it to use
370 dlopen and dlsym to refer to the original versions. This makes it possible
371 to build an LD_PRELOADable malloc replacement library.
373 THREAD_LOCAL_ALLOC Defines GC_malloc(), GC_malloc_atomic() and
374 GC_gcj_malloc() to use a per-thread set of free-lists. These then allocate
375 in a way that usually does not involve acquisition of a global lock.
376 Recommended for multiprocessors. Requires explicit GC_INIT() call, unless
377 REDIRECT_MALLOC is defined and GC_malloc is used first.
379 USE_COMPILER_TLS Causes thread local allocation to use
380 the compiler-supported "__thread" thread-local variables. This is the
381 default in HP/UX. It may help performance on recent Linux installations.
382 (It failed for me on RedHat 8, but appears to work on RedHat 9.)
384 PARALLEL_MARK Allows the marker to run in multiple threads. Recommended
387 DONT_USE_SIGNALANDWAIT (Win32 only) Use an alternate implementation for
388 marker threads (if PARALLEL_MARK defined) synchronization routines based
389 on InterlockedExchange() (instead of AO_fetch_and_add()) and on multiple
390 event objects (one per each marker instead of that based on Win32
391 SignalObjectAndWait() using a single event object). This is the default
394 GC_WINMAIN_REDIRECT (Win32 only) Redirect (rename) an application
395 WinMain to GC_WinMain; implement the "real" WinMain which starts a new
396 thread to call GC_WinMain after initializing the GC. Useful for WinCE.
397 Incompatible with GC_DLL.
399 GC_REGISTER_MEM_PRIVATE (Win32 only) Force to register MEM_PRIVATE R/W
400 sections as data roots. Might be needed for some WinCE 6.0+ custom builds.
401 (May result in numerous "Data Abort" messages logged to WinCE debugging
402 console.) Incompatible with GCC toolchains for WinCE.
404 NO_GETENV Prevents the collector from looking at environment variables.
405 These may otherwise alter its configuration, or turn off GC altogether.
406 I don't know of a reason to disable this, except possibly if the resulting
407 process runs as a privileged user. (This is on by default for WinCE.)
409 EMPTY_GETENV_RESULTS Define to workaround a reputed Wine bug in getenv
410 (getenv() may return an empty string instead of NULL for a missing entry).
412 GC_READ_ENV_FILE (Win32 only) Read environment variables from the GC "env"
413 file (named as the program name plus ".gc.env" extension). Useful for WinCE
414 targets (which have no getenv()). In the file, every variable is specified
415 in a separate line and the format is as "<name>=<value>" (without spaces).
416 A comment line may start with any character except for the Latin letters,
417 the digits and the underscore ('_'). The file encoding is Latin-1.
419 USE_GLOBAL_ALLOC (Win32 only) Use GlobalAlloc() instead of VirtualAlloc()
420 to allocate the heap. May be needed to work around a Windows NT/2000 issue.
421 Incompatible with USE_MUNMAP. See README.win32 for details.
423 MAKE_BACK_GRAPH Enable GC_PRINT_BACK_HEIGHT environment variable.
424 See README.environment for details. Experimental. Limited platform
425 support. Implies DBG_HDRS_ALL. All allocation should be done using
428 GC_PRINT_BACK_HEIGHT Permanently turn on back-height printing mode
429 (useful when NO_GETENV). See the similar environment variable description
430 in README.environment. Requires MAKE_BACK_GRAPH defined.
432 STUBBORN_ALLOC Allows allocation of "hard to change" objects, and thus
433 makes incremental collection easier. Was enabled by default until 6.0.
434 Rarely used, to my knowledge.
436 HANDLE_FORK (Unix and Cygwin only) Attempt by default to make GC_malloc()
437 work in a child process fork()'ed from a multi-threaded parent. Not fully
438 POSIX-compliant and could be disabled at runtime (before GC_INIT).
440 TEST_WITH_SYSTEM_MALLOC Causes gctest to allocate (and leak) large
441 chunks of memory with the standard system malloc. This will cause the root
442 set and collected heap to grow significantly if malloc'ed memory is somehow
443 getting traced by the collector. This has no impact on the generated
444 library; it only affects the test.
446 POINTER_MASK=<0x...> Causes candidate pointers to be AND'ed with the given
447 mask before being considered. If either this or the following macro is
448 defined, it will be assumed that all pointers stored in the heap need to be
449 processed this way. Stack and register pointers will be considered both
450 with and without processing. These macros are normally needed only to
451 support systems that use high-order pointer tags. EXPERIMENTAL.
453 POINTER_SHIFT=<n> Causes the collector to left shift candidate pointers
454 by the indicated amount before trying to interpret them. Applied after
455 POINTER_MASK. EXPERIMENTAL. See also the preceding macro.
457 ENABLE_TRACE Enables the GC_TRACE=addr environment setting to do its job.
458 By default this is not supported in order to keep the marker as fast as
461 DARWIN_DONT_PARSE_STACK Causes the Darwin port to discover thread
462 stack bounds in the same way as other pthread ports, without trying to
463 walk the frames on the stack. This is recommended only as a fall-back for
464 applications that don't support proper stack unwinding.
466 GC_NO_THREADS_DISCOVERY (Darwin and Win32+DLL only) Exclude DllMain-based
467 (on Windows) and task-threads-based (on Darwin) thread registration support.
469 GC_DISCOVER_TASK_THREADS (Darwin and Win32+DLL only) Compile the collector
470 with the implicitly turned on task-threads-based (on Darwin) or
471 DllMain-based (on Windows) approach of threads registering. Only for
472 compatibility and for the case when it is not possible to call
473 GC_use_threads_discovery() early (before other GC calls).
475 USE_PROC_FOR_LIBRARIES Causes the Linux collector to treat writable
476 memory mappings (as reported by /proc) as roots, if it doesn't have
477 other information about them. It no longer traverses dynamic loader
478 data structures to find dynamic library static data. This may be
479 required for applications that store pointers in mmapped segments without
480 informing the collector. But it typically performs poorly, especially
481 since it will scan inactive but cached NPTL thread stacks completely.
483 IGNORE_DYNAMIC_LOADING Don't define DYNAMIC_LOADING even if supported by the
484 platform (that is, build the collector with disabled tracing of dynamic
487 NO_PROC_STAT Causes the collector to avoid relying on Linux
490 NO_GETCONTEXT Causes the collector to not assume the existence of the
491 getcontext() function on linux-like platforms. This currently happens
492 implicitly on Darwin, Hurd, or ARM or MIPS hardware. It is explicitly
493 needed for some old versions of FreeBSD.
495 STATIC=static Causes various GC_ symbols that could logically be declared
496 static to be declared (this is the default if NO_DEBUGGING is specified).
497 Reduces the number of visible symbols (letting the optimizer do its work
498 better), which is probably cleaner, but may make some kinds of debugging
499 and profiling harder.
501 GC_DLL Build dynamic-link library (or dynamic shared object). For Unix this
502 causes the exported symbols to have 'default' visibility (ignored unless
503 GCC v4+) and the internal ones to have 'hidden' visibility.
505 DONT_USE_USER32_DLL (Win32 only) Don't use "user32" DLL import library
506 (containing MessageBox() entry); useful for a static GC library.
508 GC_PREFER_MPROTECT_VDB Choose MPROTECT_VDB manually in case of multiple
509 virtual dirty bit strategies are implemented (at present useful on Win32 and
510 Solaris to force MPROTECT_VDB strategy instead of the default GWW_VDB or
513 GC_IGNORE_GCJ_INFO Disable GCJ-style type information (useful for
516 GC_PRINT_VERBOSE_STATS Permanently turn on verbose logging (useful for
517 debugging and profiling on WinCE).
519 GC_ONLY_LOG_TO_FILE Don't redirect GC stdout and stderr to the log file
520 specified by GC_LOG_FILE environment variable. Has effect only when the
521 variable is set (to anything other than "0").
523 GC_DONT_EXPAND Don't expand the heap unless explicitly requested or forced to.
525 GC_USE_ENTIRE_HEAP Causes the non-incremental collector to use the
526 entire heap before collecting. This sometimes results in more large block
527 fragmentation, since very large blocks will tend to get broken up during
528 each GC cycle. It is likely to result in a larger working set, but lower
529 collection frequencies, and hence fewer instructions executed in the
530 collector. This macro controls only the default GC_use_entire_heap value.
532 GC_INITIAL_HEAP_SIZE=<value> Set the desired default initial heap size
535 GC_FREE_SPACE_DIVISOR=<value> Set alternate default GC_free_space_divisor
538 GC_TIME_LIMIT=<milliseconds> Set alternate default GC_time_limit value
539 (setting this to GC_TIME_UNLIMITED will essentially disable incremental
540 collection while leaving generational collection enabled).
542 GC_FULL_FREQ=<value> Set alternate default number of partial collections
543 between full collections (matters only if incremental collection is on).
545 NO_CANCEL_SAFE (Posix platforms with threads only) Don't bother trying
546 to make the collector safe for thread cancellation; cancellation is not
547 used. (Note that if cancellation is used anyway, threads may end up
548 getting cancelled in unexpected places.) Even without this option,
549 PTHREAD_CANCEL_ASYNCHRONOUS is never safe with the collector. (We could
550 argue about its safety without the collector.)
552 UNICODE (Win32 only) Use the Unicode variant ('W') of the Win32 API instead
553 of ANSI/ASCII one ('A'). Useful for WinCE.
555 PLATFORM_ANDROID Compile for Android NDK platform.
557 SN_TARGET_PS3 Compile for Sony PS/3.
559 USE_GET_STACKBASE_FOR_MAIN (Linux only) Use pthread_attr_getstack() instead
560 of __libc_stack_end (or instead of any hard-coded value) for getting the
561 primordial thread stack base (useful if the client modifies the program's