1 #if 0 /* Moved to malloc.h */
2 /* ---------- To make a malloc.h, start cutting here ------------ */
5 A version of malloc/free/realloc written by Doug Lea and released to the
6 public domain. Send questions/comments/complaints/performance data
9 * VERSION 2.6.6 Sun Mar 5 19:10:03 2000 Doug Lea (dl at gee)
11 Note: There may be an updated version of this malloc obtainable at
12 ftp://g.oswego.edu/pub/misc/malloc.c
13 Check before installing!
15 * Why use this malloc?
17 This is not the fastest, most space-conserving, most portable, or
18 most tunable malloc ever written. However it is among the fastest
19 while also being among the most space-conserving, portable and tunable.
20 Consistent balance across these factors results in a good general-purpose
21 allocator. For a high-level description, see
22 http://g.oswego.edu/dl/html/malloc.html
24 * Synopsis of public routines
26 (Much fuller descriptions are contained in the program documentation below.)
29 Return a pointer to a newly allocated chunk of at least n bytes, or null
30 if no space is available.
32 Release the chunk of memory pointed to by p, or no effect if p is null.
33 realloc(Void_t* p, size_t n);
34 Return a pointer to a chunk of size n that contains the same data
35 as does chunk p up to the minimum of (n, p's size) bytes, or null
36 if no space is available. The returned pointer may or may not be
37 the same as p. If p is null, equivalent to malloc. Unless the
38 #define REALLOC_ZERO_BYTES_FREES below is set, realloc with a
39 size argument of zero (re)allocates a minimum-sized chunk.
40 memalign(size_t alignment, size_t n);
41 Return a pointer to a newly allocated chunk of n bytes, aligned
42 in accord with the alignment argument, which must be a power of
45 Equivalent to memalign(pagesize, n), where pagesize is the page
46 size of the system (or as near to this as can be figured out from
47 all the includes/defines below.)
49 Equivalent to valloc(minimum-page-that-holds(n)), that is,
50 round up n to nearest pagesize.
51 calloc(size_t unit, size_t quantity);
52 Returns a pointer to quantity * unit bytes, with all locations
55 Equivalent to free(p).
56 malloc_trim(size_t pad);
57 Release all but pad bytes of freed top-most memory back
58 to the system. Return 1 if successful, else 0.
59 malloc_usable_size(Void_t* p);
60 Report the number usable allocated bytes associated with allocated
61 chunk p. This may or may not report more bytes than were requested,
62 due to alignment and minimum size constraints.
64 Prints brief summary statistics.
66 Returns (by copy) a struct containing various summary statistics.
67 mallopt(int parameter_number, int parameter_value)
68 Changes one of the tunable parameters described below. Returns
69 1 if successful in changing the parameter, else 0.
74 8 byte alignment is currently hardwired into the design. This
75 seems to suffice for all current machines and C compilers.
77 Assumed pointer representation: 4 or 8 bytes
78 Code for 8-byte pointers is untested by me but has worked
79 reliably by Wolfram Gloger, who contributed most of the
80 changes supporting this.
82 Assumed size_t representation: 4 or 8 bytes
83 Note that size_t is allowed to be 4 bytes even if pointers are 8.
85 Minimum overhead per allocated chunk: 4 or 8 bytes
86 Each malloced chunk has a hidden overhead of 4 bytes holding size
87 and status information.
89 Minimum allocated size: 4-byte ptrs: 16 bytes (including 4 overhead)
90 8-byte ptrs: 24/32 bytes (including, 4/8 overhead)
92 When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte
93 ptrs but 4 byte size) or 24 (for 8/8) additional bytes are
94 needed; 4 (8) for a trailing size field
95 and 8 (16) bytes for free list pointers. Thus, the minimum
96 allocatable size is 16/24/32 bytes.
98 Even a request for zero bytes (i.e., malloc(0)) returns a
99 pointer to something of the minimum allocatable size.
101 Maximum allocated size: 4-byte size_t: 2^31 - 8 bytes
102 8-byte size_t: 2^63 - 16 bytes
104 It is assumed that (possibly signed) size_t bit values suffice to
105 represent chunk sizes. `Possibly signed' is due to the fact
106 that `size_t' may be defined on a system as either a signed or
107 an unsigned type. To be conservative, values that would appear
108 as negative numbers are avoided.
109 Requests for sizes with a negative sign bit when the request
110 size is treaded as a long will return null.
112 Maximum overhead wastage per allocated chunk: normally 15 bytes
114 Alignnment demands, plus the minimum allocatable size restriction
115 make the normal worst-case wastage 15 bytes (i.e., up to 15
116 more bytes will be allocated than were requested in malloc), with
118 1. Because requests for zero bytes allocate non-zero space,
119 the worst case wastage for a request of zero bytes is 24 bytes.
120 2. For requests >= mmap_threshold that are serviced via
121 mmap(), the worst case wastage is 8 bytes plus the remainder
122 from a system page (the minimal mmap unit); typically 4096 bytes.
126 Here are some features that are NOT currently supported
128 * No user-definable hooks for callbacks and the like.
129 * No automated mechanism for fully checking that all accesses
130 to malloced memory stay within their bounds.
131 * No support for compaction.
133 * Synopsis of compile-time options:
135 People have reported using previous versions of this malloc on all
136 versions of Unix, sometimes by tweaking some of the defines
137 below. It has been tested most extensively on Solaris and
138 Linux. It is also reported to work on WIN32 platforms.
139 People have also reported adapting this malloc for use in
140 stand-alone embedded systems.
142 The implementation is in straight, hand-tuned ANSI C. Among other
143 consequences, it uses a lot of macros. Because of this, to be at
144 all usable, this code should be compiled using an optimizing compiler
145 (for example gcc -O2) that can simplify expressions and control
148 __STD_C (default: derived from C compiler defines)
149 Nonzero if using ANSI-standard C compiler, a C++ compiler, or
150 a C compiler sufficiently close to ANSI to get away with it.
151 DEBUG (default: NOT defined)
152 Define to enable debugging. Adds fairly extensive assertion-based
153 checking to help track down memory errors, but noticeably slows down
155 REALLOC_ZERO_BYTES_FREES (default: NOT defined)
156 Define this if you think that realloc(p, 0) should be equivalent
157 to free(p). Otherwise, since malloc returns a unique pointer for
158 malloc(0), so does realloc(p, 0).
159 HAVE_MEMCPY (default: defined)
160 Define if you are not otherwise using ANSI STD C, but still
161 have memcpy and memset in your C library and want to use them.
162 Otherwise, simple internal versions are supplied.
163 USE_MEMCPY (default: 1 if HAVE_MEMCPY is defined, 0 otherwise)
164 Define as 1 if you want the C library versions of memset and
165 memcpy called in realloc and calloc (otherwise macro versions are used).
166 At least on some platforms, the simple macro versions usually
167 outperform libc versions.
168 HAVE_MMAP (default: defined as 1)
169 Define to non-zero to optionally make malloc() use mmap() to
170 allocate very large blocks.
171 HAVE_MREMAP (default: defined as 0 unless Linux libc set)
172 Define to non-zero to optionally make realloc() use mremap() to
173 reallocate very large blocks.
174 malloc_getpagesize (default: derived from system #includes)
175 Either a constant or routine call returning the system page size.
176 HAVE_USR_INCLUDE_MALLOC_H (default: NOT defined)
177 Optionally define if you are on a system with a /usr/include/malloc.h
178 that declares struct mallinfo. It is not at all necessary to
179 define this even if you do, but will ensure consistency.
180 INTERNAL_SIZE_T (default: size_t)
181 Define to a 32-bit type (probably `unsigned int') if you are on a
182 64-bit machine, yet do not want or need to allow malloc requests of
183 greater than 2^31 to be handled. This saves space, especially for
185 INTERNAL_LINUX_C_LIB (default: NOT defined)
186 Defined only when compiled as part of Linux libc.
187 Also note that there is some odd internal name-mangling via defines
188 (for example, internally, `malloc' is named `mALLOc') needed
189 when compiling in this case. These look funny but don't otherwise
191 WIN32 (default: undefined)
192 Define this on MS win (95, nt) platforms to compile in sbrk emulation.
193 LACKS_UNISTD_H (default: undefined if not WIN32)
194 Define this if your system does not have a <unistd.h>.
195 LACKS_SYS_PARAM_H (default: undefined if not WIN32)
196 Define this if your system does not have a <sys/param.h>.
197 MORECORE (default: sbrk)
198 The name of the routine to call to obtain more memory from the system.
199 MORECORE_FAILURE (default: -1)
200 The value returned upon failure of MORECORE.
201 MORECORE_CLEARS (default 1)
202 True (1) if the routine mapped to MORECORE zeroes out memory (which
204 DEFAULT_TRIM_THRESHOLD
206 DEFAULT_MMAP_THRESHOLD
208 Default values of tunable parameters (described in detail below)
209 controlling interaction with host system routines (sbrk, mmap, etc).
210 These values may also be changed dynamically via mallopt(). The
211 preset defaults are those that give best performance for typical
213 USE_DL_PREFIX (default: undefined)
214 Prefix all public routines with the string 'dl'. Useful to
215 quickly avoid procedure declaration conflicts and linker symbol
216 conflicts with existing memory allocation routines.
234 #endif /*__cplusplus*/
239 #if (__STD_C || defined(WIN32))
247 #include <stddef.h> /* for size_t */
249 #include <sys/types.h>
256 #include <stdio.h> /* needed for malloc_stats */
267 Because freed chunks may be overwritten with link fields, this
268 malloc will often die when freed memory is overwritten by user
269 programs. This can be very effective (albeit in an annoying way)
270 in helping track down dangling pointers.
272 If you compile with -DDEBUG, a number of assertion checks are
273 enabled that will catch more memory errors. You probably won't be
274 able to make much sense of the actual assertion errors, but they
275 should help you locate incorrectly overwritten memory. The
276 checking is fairly extensive, and will slow down execution
277 noticeably. Calling malloc_stats or mallinfo with DEBUG set will
278 attempt to check every non-mmapped allocated and free chunk in the
279 course of computing the summmaries. (By nature, mmapped regions
280 cannot be checked very much automatically.)
282 Setting DEBUG may also be helpful if you are trying to modify
283 this code. The assertions in the check routines spell out in more
284 detail the assumptions and invariants underlying the algorithms.
291 #define assert(x) ((void)0)
296 INTERNAL_SIZE_T is the word-size used for internal bookkeeping
297 of chunk sizes. On a 64-bit machine, you can reduce malloc
298 overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
299 at the expense of not being able to handle requests greater than
300 2^31. This limitation is hardly ever a concern; you are encouraged
301 to set this. However, the default version is the same as size_t.
304 #ifndef INTERNAL_SIZE_T
305 #define INTERNAL_SIZE_T size_t
309 REALLOC_ZERO_BYTES_FREES should be set if a call to
310 realloc with zero bytes should be the same as a call to free.
311 Some people think it should. Otherwise, since this malloc
312 returns a unique pointer for malloc(0), so does realloc(p, 0).
316 /* #define REALLOC_ZERO_BYTES_FREES */
320 WIN32 causes an emulation of sbrk to be compiled in
321 mmap-based options are not currently supported in WIN32.
326 #define MORECORE wsbrk
329 #define LACKS_UNISTD_H
330 #define LACKS_SYS_PARAM_H
333 Include 'windows.h' to get the necessary declarations for the
334 Microsoft Visual C++ data structures and routines used in the 'sbrk'
337 Define WIN32_LEAN_AND_MEAN so that only the essential Microsoft
338 Visual C++ header files are included.
340 #define WIN32_LEAN_AND_MEAN
346 HAVE_MEMCPY should be defined if you are not otherwise using
347 ANSI STD C, but still have memcpy and memset in your C library
348 and want to use them in calloc and realloc. Otherwise simple
349 macro versions are defined here.
351 USE_MEMCPY should be defined as 1 if you actually want to
352 have memset and memcpy called. People report that the macro
353 versions are often enough faster than libc versions on many
354 systems that it is better to use them.
368 #if (__STD_C || defined(HAVE_MEMCPY))
371 void* memset(void*, int, size_t);
372 void* memcpy(void*, const void*, size_t);
375 // On Win32 platforms, 'memset()' and 'memcpy()' are already declared in
386 /* The following macros are only invoked with (2n+1)-multiples of
387 INTERNAL_SIZE_T units, with a positive integer n. This is exploited
388 for fast inline execution when n is small. */
390 #define MALLOC_ZERO(charp, nbytes) \
392 INTERNAL_SIZE_T mzsz = (nbytes); \
393 if(mzsz <= 9*sizeof(mzsz)) { \
394 INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp); \
395 if(mzsz >= 5*sizeof(mzsz)) { *mz++ = 0; \
397 if(mzsz >= 7*sizeof(mzsz)) { *mz++ = 0; \
399 if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0; \
404 } else memset((charp), 0, mzsz); \
407 #define MALLOC_COPY(dest,src,nbytes) \
409 INTERNAL_SIZE_T mcsz = (nbytes); \
410 if(mcsz <= 9*sizeof(mcsz)) { \
411 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src); \
412 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest); \
413 if(mcsz >= 5*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
414 *mcdst++ = *mcsrc++; \
415 if(mcsz >= 7*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
416 *mcdst++ = *mcsrc++; \
417 if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
418 *mcdst++ = *mcsrc++; }}} \
419 *mcdst++ = *mcsrc++; \
420 *mcdst++ = *mcsrc++; \
422 } else memcpy(dest, src, mcsz); \
425 #else /* !USE_MEMCPY */
427 /* Use Duff's device for good zeroing/copying performance. */
429 #define MALLOC_ZERO(charp, nbytes) \
431 INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp); \
432 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
433 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
435 case 0: for(;;) { *mzp++ = 0; \
436 case 7: *mzp++ = 0; \
437 case 6: *mzp++ = 0; \
438 case 5: *mzp++ = 0; \
439 case 4: *mzp++ = 0; \
440 case 3: *mzp++ = 0; \
441 case 2: *mzp++ = 0; \
442 case 1: *mzp++ = 0; if(mcn <= 0) break; mcn--; } \
446 #define MALLOC_COPY(dest,src,nbytes) \
448 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src; \
449 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest; \
450 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
451 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
453 case 0: for(;;) { *mcdst++ = *mcsrc++; \
454 case 7: *mcdst++ = *mcsrc++; \
455 case 6: *mcdst++ = *mcsrc++; \
456 case 5: *mcdst++ = *mcsrc++; \
457 case 4: *mcdst++ = *mcsrc++; \
458 case 3: *mcdst++ = *mcsrc++; \
459 case 2: *mcdst++ = *mcsrc++; \
460 case 1: *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; } \
468 Define HAVE_MMAP to optionally make malloc() use mmap() to
469 allocate very large blocks. These will be returned to the
470 operating system immediately after a free().
478 Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
479 large blocks. This is currently only possible on Linux with
480 kernel versions newer than 1.3.77.
484 #ifdef INTERNAL_LINUX_C_LIB
485 #define HAVE_MREMAP 1
487 #define HAVE_MREMAP 0
495 #include <sys/mman.h>
497 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
498 #define MAP_ANONYMOUS MAP_ANON
501 #endif /* HAVE_MMAP */
504 Access to system page size. To the extent possible, this malloc
505 manages memory from the system in page-size units.
507 The following mechanics for getpagesize were adapted from
508 bsd/gnu getpagesize.h
511 #ifndef LACKS_UNISTD_H
515 #ifndef malloc_getpagesize
516 # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
517 # ifndef _SC_PAGE_SIZE
518 # define _SC_PAGE_SIZE _SC_PAGESIZE
521 # ifdef _SC_PAGE_SIZE
522 # define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
524 # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
525 extern size_t getpagesize();
526 # define malloc_getpagesize getpagesize()
529 # define malloc_getpagesize (4096) /* TBD: Use 'GetSystemInfo' instead */
531 # ifndef LACKS_SYS_PARAM_H
532 # include <sys/param.h>
534 # ifdef EXEC_PAGESIZE
535 # define malloc_getpagesize EXEC_PAGESIZE
539 # define malloc_getpagesize NBPG
541 # define malloc_getpagesize (NBPG * CLSIZE)
545 # define malloc_getpagesize NBPC
548 # define malloc_getpagesize PAGESIZE
550 # define malloc_getpagesize (4096) /* just guess */
564 This version of malloc supports the standard SVID/XPG mallinfo
565 routine that returns a struct containing the same kind of
566 information you can get from malloc_stats. It should work on
567 any SVID/XPG compliant system that has a /usr/include/malloc.h
568 defining struct mallinfo. (If you'd like to install such a thing
569 yourself, cut out the preliminary declarations as described above
570 and below and save them in a malloc.h file. But there's no
571 compelling reason to bother to do this.)
573 The main declaration needed is the mallinfo struct that is returned
574 (by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a
575 bunch of fields, most of which are not even meaningful in this
576 version of malloc. Some of these fields are are instead filled by
577 mallinfo() with other numbers that might possibly be of interest.
579 HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
580 /usr/include/malloc.h file that includes a declaration of struct
581 mallinfo. If so, it is included; else an SVID2/XPG2 compliant
582 version is declared below. These must be precisely the same for
587 /* #define HAVE_USR_INCLUDE_MALLOC_H */
589 #if HAVE_USR_INCLUDE_MALLOC_H
590 #include "/usr/include/malloc.h"
593 /* SVID2/XPG mallinfo structure */
596 int arena; /* total space allocated from system */
597 int ordblks; /* number of non-inuse chunks */
598 int smblks; /* unused -- always zero */
599 int hblks; /* number of mmapped regions */
600 int hblkhd; /* total space in mmapped regions */
601 int usmblks; /* unused -- always zero */
602 int fsmblks; /* unused -- always zero */
603 int uordblks; /* total allocated space */
604 int fordblks; /* total non-inuse space */
605 int keepcost; /* top-most, releasable (via malloc_trim) space */
608 /* SVID2/XPG mallopt options */
610 #define M_MXFAST 1 /* UNUSED in this malloc */
611 #define M_NLBLKS 2 /* UNUSED in this malloc */
612 #define M_GRAIN 3 /* UNUSED in this malloc */
613 #define M_KEEP 4 /* UNUSED in this malloc */
617 /* mallopt options that actually do something */
619 #define M_TRIM_THRESHOLD -1
621 #define M_MMAP_THRESHOLD -3
622 #define M_MMAP_MAX -4
626 #ifndef DEFAULT_TRIM_THRESHOLD
627 #define DEFAULT_TRIM_THRESHOLD (128 * 1024)
631 M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
632 to keep before releasing via malloc_trim in free().
634 Automatic trimming is mainly useful in long-lived programs.
635 Because trimming via sbrk can be slow on some systems, and can
636 sometimes be wasteful (in cases where programs immediately
637 afterward allocate more large chunks) the value should be high
638 enough so that your overall system performance would improve by
641 The trim threshold and the mmap control parameters (see below)
642 can be traded off with one another. Trimming and mmapping are
643 two different ways of releasing unused memory back to the
644 system. Between these two, it is often possible to keep
645 system-level demands of a long-lived program down to a bare
646 minimum. For example, in one test suite of sessions measuring
647 the XF86 X server on Linux, using a trim threshold of 128K and a
648 mmap threshold of 192K led to near-minimal long term resource
651 If you are using this malloc in a long-lived program, it should
652 pay to experiment with these values. As a rough guide, you
653 might set to a value close to the average size of a process
654 (program) running on your system. Releasing this much memory
655 would allow such a process to run in memory. Generally, it's
656 worth it to tune for trimming rather tham memory mapping when a
657 program undergoes phases where several large chunks are
658 allocated and released in ways that can reuse each other's
659 storage, perhaps mixed with phases where there are no such
660 chunks at all. And in well-behaved long-lived programs,
661 controlling release of large blocks via trimming versus mapping
664 However, in most programs, these parameters serve mainly as
665 protection against the system-level effects of carrying around
666 massive amounts of unneeded memory. Since frequent calls to
667 sbrk, mmap, and munmap otherwise degrade performance, the default
668 parameters are set to relatively high values that serve only as
671 The default trim value is high enough to cause trimming only in
672 fairly extreme (by current memory consumption standards) cases.
673 It must be greater than page size to have any useful effect. To
674 disable trimming completely, you can set to (unsigned long)(-1);
680 #ifndef DEFAULT_TOP_PAD
681 #define DEFAULT_TOP_PAD (0)
685 M_TOP_PAD is the amount of extra `padding' space to allocate or
686 retain whenever sbrk is called. It is used in two ways internally:
688 * When sbrk is called to extend the top of the arena to satisfy
689 a new malloc request, this much padding is added to the sbrk
692 * When malloc_trim is called automatically from free(),
693 it is used as the `pad' argument.
695 In both cases, the actual amount of padding is rounded
696 so that the end of the arena is always a system page boundary.
698 The main reason for using padding is to avoid calling sbrk so
699 often. Having even a small pad greatly reduces the likelihood
700 that nearly every malloc request during program start-up (or
701 after trimming) will invoke sbrk, which needlessly wastes
704 Automatic rounding-up to page-size units is normally sufficient
705 to avoid measurable overhead, so the default is 0. However, in
706 systems where sbrk is relatively slow, it can pay to increase
707 this value, at the expense of carrying around more memory than
713 #ifndef DEFAULT_MMAP_THRESHOLD
714 #define DEFAULT_MMAP_THRESHOLD (128 * 1024)
719 M_MMAP_THRESHOLD is the request size threshold for using mmap()
720 to service a request. Requests of at least this size that cannot
721 be allocated using already-existing space will be serviced via mmap.
722 (If enough normal freed space already exists it is used instead.)
724 Using mmap segregates relatively large chunks of memory so that
725 they can be individually obtained and released from the host
726 system. A request serviced through mmap is never reused by any
727 other request (at least not directly; the system may just so
728 happen to remap successive requests to the same locations).
730 Segregating space in this way has the benefit that mmapped space
731 can ALWAYS be individually released back to the system, which
732 helps keep the system level memory demands of a long-lived
733 program low. Mapped memory can never become `locked' between
734 other chunks, as can happen with normally allocated chunks, which
735 menas that even trimming via malloc_trim would not release them.
737 However, it has the disadvantages that:
739 1. The space cannot be reclaimed, consolidated, and then
740 used to service later requests, as happens with normal chunks.
741 2. It can lead to more wastage because of mmap page alignment
743 3. It causes malloc performance to be more dependent on host
744 system memory management support routines which may vary in
745 implementation quality and may impose arbitrary
746 limitations. Generally, servicing a request via normal
747 malloc steps is faster than going through a system's mmap.
749 All together, these considerations should lead you to use mmap
750 only for relatively large requests.
757 #ifndef DEFAULT_MMAP_MAX
759 #define DEFAULT_MMAP_MAX (64)
761 #define DEFAULT_MMAP_MAX (0)
766 M_MMAP_MAX is the maximum number of requests to simultaneously
767 service using mmap. This parameter exists because:
769 1. Some systems have a limited number of internal tables for
771 2. In most systems, overreliance on mmap can degrade overall
773 3. If a program allocates many large regions, it is probably
774 better off using normal sbrk-based allocation routines that
775 can reclaim and reallocate normal heap memory. Using a
776 small value allows transition into this mode after the
777 first few allocations.
779 Setting to 0 disables all use of mmap. If HAVE_MMAP is not set,
780 the default value is 0, and attempts to set it to non-zero values
781 in mallopt will fail.
788 USE_DL_PREFIX will prefix all public routines with the string 'dl'.
789 Useful to quickly avoid procedure declaration conflicts and linker
790 symbol conflicts with existing memory allocation routines.
794 /* #define USE_DL_PREFIX */
801 Special defines for linux libc
803 Except when compiled using these special defines for Linux libc
804 using weak aliases, this malloc is NOT designed to work in
805 multithreaded applications. No semaphores or other concurrency
806 control are provided to ensure that multiple malloc or free calls
807 don't run at the same time, which could be disasterous. A single
808 semaphore could be used across malloc, realloc, and free (which is
809 essentially the effect of the linux weak alias approach). It would
810 be hard to obtain finer granularity.
815 #ifdef INTERNAL_LINUX_C_LIB
819 Void_t * __default_morecore_init (ptrdiff_t);
820 Void_t *(*__morecore)(ptrdiff_t) = __default_morecore_init;
824 Void_t * __default_morecore_init ();
825 Void_t *(*__morecore)() = __default_morecore_init;
829 #define MORECORE (*__morecore)
830 #define MORECORE_FAILURE 0
831 #define MORECORE_CLEARS 1
833 #else /* INTERNAL_LINUX_C_LIB */
836 extern Void_t* sbrk(ptrdiff_t);
838 extern Void_t* sbrk();
842 #define MORECORE sbrk
845 #ifndef MORECORE_FAILURE
846 #define MORECORE_FAILURE -1
849 #ifndef MORECORE_CLEARS
850 #define MORECORE_CLEARS 1
853 #endif /* INTERNAL_LINUX_C_LIB */
855 #if defined(INTERNAL_LINUX_C_LIB) && defined(__ELF__)
857 #define cALLOc __libc_calloc
858 #define fREe __libc_free
859 #define mALLOc __libc_malloc
860 #define mEMALIGn __libc_memalign
861 #define rEALLOc __libc_realloc
862 #define vALLOc __libc_valloc
863 #define pvALLOc __libc_pvalloc
864 #define mALLINFo __libc_mallinfo
865 #define mALLOPt __libc_mallopt
867 #pragma weak calloc = __libc_calloc
868 #pragma weak free = __libc_free
869 #pragma weak cfree = __libc_free
870 #pragma weak malloc = __libc_malloc
871 #pragma weak memalign = __libc_memalign
872 #pragma weak realloc = __libc_realloc
873 #pragma weak valloc = __libc_valloc
874 #pragma weak pvalloc = __libc_pvalloc
875 #pragma weak mallinfo = __libc_mallinfo
876 #pragma weak mallopt = __libc_mallopt
881 #define cALLOc dlcalloc
883 #define mALLOc dlmalloc
884 #define mEMALIGn dlmemalign
885 #define rEALLOc dlrealloc
886 #define vALLOc dlvalloc
887 #define pvALLOc dlpvalloc
888 #define mALLINFo dlmallinfo
889 #define mALLOPt dlmallopt
890 #else /* USE_DL_PREFIX */
891 #define cALLOc calloc
893 #define mALLOc malloc
894 #define mEMALIGn memalign
895 #define rEALLOc realloc
896 #define vALLOc valloc
897 #define pvALLOc pvalloc
898 #define mALLINFo mallinfo
899 #define mALLOPt mallopt
900 #endif /* USE_DL_PREFIX */
904 /* Public routines */
908 Void_t* mALLOc(size_t);
910 Void_t* rEALLOc(Void_t*, size_t);
911 Void_t* mEMALIGn(size_t, size_t);
912 Void_t* vALLOc(size_t);
913 Void_t* pvALLOc(size_t);
914 Void_t* cALLOc(size_t, size_t);
916 int malloc_trim(size_t);
917 size_t malloc_usable_size(Void_t*);
919 int mALLOPt(int, int);
920 struct mallinfo mALLINFo(void);
931 size_t malloc_usable_size();
934 struct mallinfo mALLINFo();
939 }; /* end of extern "C" */
942 /* ---------- To make a malloc.h, end cutting here ------------ */
943 #else /* Moved to malloc.h */
948 static void malloc_update_mallinfo (void);
949 void malloc_stats (void);
951 static void malloc_update_mallinfo ();
956 #endif /* 0 */ /* Moved to malloc.h */
960 Emulation of sbrk for WIN32
961 All code within the ifdef WIN32 is untested by me.
963 Thanks to Martin Fong and others for supplying this.
969 #define AlignPage(add) (((add) + (malloc_getpagesize-1)) & \
970 ~(malloc_getpagesize-1))
971 #define AlignPage64K(add) (((add) + (0x10000 - 1)) & ~(0x10000 - 1))
973 /* resrve 64MB to insure large contiguous space */
974 #define RESERVED_SIZE (1024*1024*64)
975 #define NEXT_SIZE (2048*1024)
976 #define TOP_MEMORY ((unsigned long)2*1024*1024*1024)
978 struct GmListElement;
979 typedef struct GmListElement GmListElement;
987 static GmListElement* head = 0;
988 static unsigned int gNextAddress = 0;
989 static unsigned int gAddressBase = 0;
990 static unsigned int gAllocatedSize = 0;
993 GmListElement* makeGmListElement (void* bas)
996 this = (GmListElement*)(void*)LocalAlloc (0, sizeof (GmListElement));
1010 assert ( (head == NULL) || (head->base == (void*)gAddressBase));
1011 if (gAddressBase && (gNextAddress - gAddressBase))
1013 rval = VirtualFree ((void*)gAddressBase,
1014 gNextAddress - gAddressBase,
1020 GmListElement* next = head->next;
1021 rval = VirtualFree (head->base, 0, MEM_RELEASE);
1029 void* findRegion (void* start_address, unsigned long size)
1031 MEMORY_BASIC_INFORMATION info;
1032 if (size >= TOP_MEMORY) return NULL;
1034 while ((unsigned long)start_address + size < TOP_MEMORY)
1036 VirtualQuery (start_address, &info, sizeof (info));
1037 if ((info.State == MEM_FREE) && (info.RegionSize >= size))
1038 return start_address;
1041 // Requested region is not available so see if the
1042 // next region is available. Set 'start_address'
1043 // to the next region and call 'VirtualQuery()'
1046 start_address = (char*)info.BaseAddress + info.RegionSize;
1048 // Make sure we start looking for the next region
1049 // on the *next* 64K boundary. Otherwise, even if
1050 // the new region is free according to
1051 // 'VirtualQuery()', the subsequent call to
1052 // 'VirtualAlloc()' (which follows the call to
1053 // this routine in 'wsbrk()') will round *down*
1054 // the requested address to a 64K boundary which
1055 // we already know is an address in the
1056 // unavailable region. Thus, the subsequent call
1057 // to 'VirtualAlloc()' will fail and bring us back
1058 // here, causing us to go into an infinite loop.
1061 (void *) AlignPage64K((unsigned long) start_address);
1069 void* wsbrk (long size)
1074 if (gAddressBase == 0)
1076 gAllocatedSize = max (RESERVED_SIZE, AlignPage (size));
1077 gNextAddress = gAddressBase =
1078 (unsigned int)VirtualAlloc (NULL, gAllocatedSize,
1079 MEM_RESERVE, PAGE_NOACCESS);
1080 } else if (AlignPage (gNextAddress + size) > (gAddressBase +
1083 long new_size = max (NEXT_SIZE, AlignPage (size));
1084 void* new_address = (void*)(gAddressBase+gAllocatedSize);
1087 new_address = findRegion (new_address, new_size);
1089 if (new_address == 0)
1092 gAddressBase = gNextAddress =
1093 (unsigned int)VirtualAlloc (new_address, new_size,
1094 MEM_RESERVE, PAGE_NOACCESS);
1095 // repeat in case of race condition
1096 // The region that we found has been snagged
1097 // by another thread
1099 while (gAddressBase == 0);
1101 assert (new_address == (void*)gAddressBase);
1103 gAllocatedSize = new_size;
1105 if (!makeGmListElement ((void*)gAddressBase))
1108 if ((size + gNextAddress) > AlignPage (gNextAddress))
1111 res = VirtualAlloc ((void*)AlignPage (gNextAddress),
1112 (size + gNextAddress -
1113 AlignPage (gNextAddress)),
1114 MEM_COMMIT, PAGE_READWRITE);
1118 tmp = (void*)gNextAddress;
1119 gNextAddress = (unsigned int)tmp + size;
1124 unsigned int alignedGoal = AlignPage (gNextAddress + size);
1125 /* Trim by releasing the virtual memory */
1126 if (alignedGoal >= gAddressBase)
1128 VirtualFree ((void*)alignedGoal, gNextAddress - alignedGoal,
1130 gNextAddress = gNextAddress + size;
1131 return (void*)gNextAddress;
1135 VirtualFree ((void*)gAddressBase, gNextAddress - gAddressBase,
1137 gNextAddress = gAddressBase;
1143 return (void*)gNextAddress;
1158 INTERNAL_SIZE_T prev_size; /* Size of previous chunk (if free). */
1159 INTERNAL_SIZE_T size; /* Size in bytes, including overhead. */
1160 struct malloc_chunk* fd; /* double links -- used only if free. */
1161 struct malloc_chunk* bk;
1164 typedef struct malloc_chunk* mchunkptr;
1168 malloc_chunk details:
1170 (The following includes lightly edited explanations by Colin Plumb.)
1172 Chunks of memory are maintained using a `boundary tag' method as
1173 described in e.g., Knuth or Standish. (See the paper by Paul
1174 Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
1175 survey of such techniques.) Sizes of free chunks are stored both
1176 in the front of each chunk and at the end. This makes
1177 consolidating fragmented chunks into bigger chunks very fast. The
1178 size fields also hold bits representing whether chunks are free or
1181 An allocated chunk looks like this:
1184 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1185 | Size of previous chunk, if allocated | |
1186 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1187 | Size of chunk, in bytes |P|
1188 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1189 | User data starts here... .
1191 . (malloc_usable_space() bytes) .
1193 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1195 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1198 Where "chunk" is the front of the chunk for the purpose of most of
1199 the malloc code, but "mem" is the pointer that is returned to the
1200 user. "Nextchunk" is the beginning of the next contiguous chunk.
1202 Chunks always begin on even word boundries, so the mem portion
1203 (which is returned to the user) is also on an even word boundary, and
1204 thus double-word aligned.
1206 Free chunks are stored in circular doubly-linked lists, and look like this:
1208 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1209 | Size of previous chunk |
1210 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1211 `head:' | Size of chunk, in bytes |P|
1212 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1213 | Forward pointer to next chunk in list |
1214 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1215 | Back pointer to previous chunk in list |
1216 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1217 | Unused space (may be 0 bytes long) .
1220 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1221 `foot:' | Size of chunk, in bytes |
1222 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1224 The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1225 chunk size (which is always a multiple of two words), is an in-use
1226 bit for the *previous* chunk. If that bit is *clear*, then the
1227 word before the current chunk size contains the previous chunk
1228 size, and can be used to find the front of the previous chunk.
1229 (The very first chunk allocated always has this bit set,
1230 preventing access to non-existent (or non-owned) memory.)
1232 Note that the `foot' of the current chunk is actually represented
1233 as the prev_size of the NEXT chunk. (This makes it easier to
1234 deal with alignments etc).
1236 The two exceptions to all this are
1238 1. The special chunk `top', which doesn't bother using the
1239 trailing size field since there is no
1240 next contiguous chunk that would have to index off it. (After
1241 initialization, `top' is forced to always exist. If it would
1242 become less than MINSIZE bytes long, it is replenished via
1245 2. Chunks allocated via mmap, which have the second-lowest-order
1246 bit (IS_MMAPPED) set in their size fields. Because they are
1247 never merged or traversed from any other chunk, they have no
1248 foot size or inuse information.
1250 Available chunks are kept in any of several places (all declared below):
1252 * `av': An array of chunks serving as bin headers for consolidated
1253 chunks. Each bin is doubly linked. The bins are approximately
1254 proportionally (log) spaced. There are a lot of these bins
1255 (128). This may look excessive, but works very well in
1256 practice. All procedures maintain the invariant that no
1257 consolidated chunk physically borders another one. Chunks in
1258 bins are kept in size order, with ties going to the
1259 approximately least recently used chunk.
1261 The chunks in each bin are maintained in decreasing sorted order by
1262 size. This is irrelevant for the small bins, which all contain
1263 the same-sized chunks, but facilitates best-fit allocation for
1264 larger chunks. (These lists are just sequential. Keeping them in
1265 order almost never requires enough traversal to warrant using
1266 fancier ordered data structures.) Chunks of the same size are
1267 linked with the most recently freed at the front, and allocations
1268 are taken from the back. This results in LRU or FIFO allocation
1269 order, which tends to give each chunk an equal opportunity to be
1270 consolidated with adjacent freed chunks, resulting in larger free
1271 chunks and less fragmentation.
1273 * `top': The top-most available chunk (i.e., the one bordering the
1274 end of available memory) is treated specially. It is never
1275 included in any bin, is used only if no other chunk is
1276 available, and is released back to the system if it is very
1277 large (see M_TRIM_THRESHOLD).
1279 * `last_remainder': A bin holding only the remainder of the
1280 most recently split (non-top) chunk. This bin is checked
1281 before other non-fitting chunks, so as to provide better
1282 locality for runs of sequentially allocated chunks.
1284 * Implicitly, through the host system's memory mapping tables.
1285 If supported, requests greater than a threshold are usually
1286 serviced via calls to mmap, and then later released via munmap.
1295 /* sizes, alignments */
1297 #define SIZE_SZ (sizeof(INTERNAL_SIZE_T))
1298 #define MALLOC_ALIGNMENT (SIZE_SZ + SIZE_SZ)
1299 #define MALLOC_ALIGN_MASK (MALLOC_ALIGNMENT - 1)
1300 #define MINSIZE (sizeof(struct malloc_chunk))
1302 /* conversion from malloc headers to user pointers, and back */
1304 #define chunk2mem(p) ((Void_t*)((char*)(p) + 2*SIZE_SZ))
1305 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1307 /* pad request bytes into a usable size */
1309 #define request2size(req) \
1310 (((long)((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) < \
1311 (long)(MINSIZE + MALLOC_ALIGN_MASK)) ? MINSIZE : \
1312 (((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) & ~(MALLOC_ALIGN_MASK)))
1314 /* Check if m has acceptable alignment */
1316 #define aligned_OK(m) (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
1322 Physical chunk operations
1326 /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1328 #define PREV_INUSE 0x1
1330 /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1332 #define IS_MMAPPED 0x2
1334 /* Bits to mask off when extracting size */
1336 #define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
1339 /* Ptr to next physical malloc_chunk. */
1341 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & ~PREV_INUSE) ))
1343 /* Ptr to previous physical malloc_chunk */
1345 #define prev_chunk(p)\
1346 ((mchunkptr)( ((char*)(p)) - ((p)->prev_size) ))
1349 /* Treat space at ptr + offset as a chunk */
1351 #define chunk_at_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
1357 Dealing with use bits
1360 /* extract p's inuse bit */
1363 ((((mchunkptr)(((char*)(p))+((p)->size & ~PREV_INUSE)))->size) & PREV_INUSE)
1365 /* extract inuse bit of previous chunk */
1367 #define prev_inuse(p) ((p)->size & PREV_INUSE)
1369 /* check for mmap()'ed chunk */
1371 #define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
1373 /* set/clear chunk as in use without otherwise disturbing */
1375 #define set_inuse(p)\
1376 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size |= PREV_INUSE
1378 #define clear_inuse(p)\
1379 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size &= ~(PREV_INUSE)
1381 /* check/set/clear inuse bits in known places */
1383 #define inuse_bit_at_offset(p, s)\
1384 (((mchunkptr)(((char*)(p)) + (s)))->size & PREV_INUSE)
1386 #define set_inuse_bit_at_offset(p, s)\
1387 (((mchunkptr)(((char*)(p)) + (s)))->size |= PREV_INUSE)
1389 #define clear_inuse_bit_at_offset(p, s)\
1390 (((mchunkptr)(((char*)(p)) + (s)))->size &= ~(PREV_INUSE))
1396 Dealing with size fields
1399 /* Get size, ignoring use bits */
1401 #define chunksize(p) ((p)->size & ~(SIZE_BITS))
1403 /* Set size at head, without disturbing its use bit */
1405 #define set_head_size(p, s) ((p)->size = (((p)->size & PREV_INUSE) | (s)))
1407 /* Set size/use ignoring previous bits in header */
1409 #define set_head(p, s) ((p)->size = (s))
1411 /* Set size at footer (only when chunk is not in use) */
1413 #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_size = (s))
1422 The bins, `av_' are an array of pairs of pointers serving as the
1423 heads of (initially empty) doubly-linked lists of chunks, laid out
1424 in a way so that each pair can be treated as if it were in a
1425 malloc_chunk. (This way, the fd/bk offsets for linking bin heads
1426 and chunks are the same).
1428 Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1429 8 bytes apart. Larger bins are approximately logarithmically
1430 spaced. (See the table below.) The `av_' array is never mentioned
1431 directly in the code, but instead via bin access macros.
1439 4 bins of size 32768
1440 2 bins of size 262144
1441 1 bin of size what's left
1443 There is actually a little bit of slop in the numbers in bin_index
1444 for the sake of speed. This makes no difference elsewhere.
1446 The special chunks `top' and `last_remainder' get their own bins,
1447 (this is implemented via yet more trickery with the av_ array),
1448 although `top' is never properly linked to its bin since it is
1449 always handled specially.
1453 #define NAV 128 /* number of bins */
1455 typedef struct malloc_chunk* mbinptr;
1459 #define bin_at(i) ((mbinptr)((char*)&(av_[2*(i) + 2]) - 2*SIZE_SZ))
1460 #define next_bin(b) ((mbinptr)((char*)(b) + 2 * sizeof(mbinptr)))
1461 #define prev_bin(b) ((mbinptr)((char*)(b) - 2 * sizeof(mbinptr)))
1464 The first 2 bins are never indexed. The corresponding av_ cells are instead
1465 used for bookkeeping. This is not to save space, but to simplify
1466 indexing, maintain locality, and avoid some initialization tests.
1469 #define top (bin_at(0)->fd) /* The topmost chunk */
1470 #define last_remainder (bin_at(1)) /* remainder from last split */
1474 Because top initially points to its own bin with initial
1475 zero size, thus forcing extension on the first malloc request,
1476 we avoid having any special code in malloc to check whether
1477 it even exists yet. But we still need to in malloc_extend_top.
1480 #define initial_top ((mchunkptr)(bin_at(0)))
1482 /* Helper macro to initialize bins */
1484 #define IAV(i) bin_at(i), bin_at(i)
1486 static mbinptr av_[NAV * 2 + 2] = {
1488 IAV(0), IAV(1), IAV(2), IAV(3), IAV(4), IAV(5), IAV(6), IAV(7),
1489 IAV(8), IAV(9), IAV(10), IAV(11), IAV(12), IAV(13), IAV(14), IAV(15),
1490 IAV(16), IAV(17), IAV(18), IAV(19), IAV(20), IAV(21), IAV(22), IAV(23),
1491 IAV(24), IAV(25), IAV(26), IAV(27), IAV(28), IAV(29), IAV(30), IAV(31),
1492 IAV(32), IAV(33), IAV(34), IAV(35), IAV(36), IAV(37), IAV(38), IAV(39),
1493 IAV(40), IAV(41), IAV(42), IAV(43), IAV(44), IAV(45), IAV(46), IAV(47),
1494 IAV(48), IAV(49), IAV(50), IAV(51), IAV(52), IAV(53), IAV(54), IAV(55),
1495 IAV(56), IAV(57), IAV(58), IAV(59), IAV(60), IAV(61), IAV(62), IAV(63),
1496 IAV(64), IAV(65), IAV(66), IAV(67), IAV(68), IAV(69), IAV(70), IAV(71),
1497 IAV(72), IAV(73), IAV(74), IAV(75), IAV(76), IAV(77), IAV(78), IAV(79),
1498 IAV(80), IAV(81), IAV(82), IAV(83), IAV(84), IAV(85), IAV(86), IAV(87),
1499 IAV(88), IAV(89), IAV(90), IAV(91), IAV(92), IAV(93), IAV(94), IAV(95),
1500 IAV(96), IAV(97), IAV(98), IAV(99), IAV(100), IAV(101), IAV(102), IAV(103),
1501 IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
1502 IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
1503 IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
1506 void malloc_bin_reloc (void)
1508 DECLARE_GLOBAL_DATA_PTR;
1510 unsigned long *p = (unsigned long *)(&av_[2]);
1512 for (i=2; i<(sizeof(av_)/sizeof(mbinptr)); ++i) {
1513 *p++ += gd->reloc_off;
1518 /* field-extraction macros */
1520 #define first(b) ((b)->fd)
1521 #define last(b) ((b)->bk)
1527 #define bin_index(sz) \
1528 (((((unsigned long)(sz)) >> 9) == 0) ? (((unsigned long)(sz)) >> 3): \
1529 ((((unsigned long)(sz)) >> 9) <= 4) ? 56 + (((unsigned long)(sz)) >> 6): \
1530 ((((unsigned long)(sz)) >> 9) <= 20) ? 91 + (((unsigned long)(sz)) >> 9): \
1531 ((((unsigned long)(sz)) >> 9) <= 84) ? 110 + (((unsigned long)(sz)) >> 12): \
1532 ((((unsigned long)(sz)) >> 9) <= 340) ? 119 + (((unsigned long)(sz)) >> 15): \
1533 ((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18): \
1536 bins for chunks < 512 are all spaced 8 bytes apart, and hold
1537 identically sized chunks. This is exploited in malloc.
1540 #define MAX_SMALLBIN 63
1541 #define MAX_SMALLBIN_SIZE 512
1542 #define SMALLBIN_WIDTH 8
1544 #define smallbin_index(sz) (((unsigned long)(sz)) >> 3)
1547 Requests are `small' if both the corresponding and the next bin are small
1550 #define is_small_request(nb) (nb < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
1555 To help compensate for the large number of bins, a one-level index
1556 structure is used for bin-by-bin searching. `binblocks' is a
1557 one-word bitvector recording whether groups of BINBLOCKWIDTH bins
1558 have any (possibly) non-empty bins, so they can be skipped over
1559 all at once during during traversals. The bits are NOT always
1560 cleared as soon as all bins in a block are empty, but instead only
1561 when all are noticed to be empty during traversal in malloc.
1564 #define BINBLOCKWIDTH 4 /* bins per block */
1566 #define binblocks (bin_at(0)->size) /* bitvector of nonempty blocks */
1568 /* bin<->block macros */
1570 #define idx2binblock(ix) ((unsigned)1 << (ix / BINBLOCKWIDTH))
1571 #define mark_binblock(ii) (binblocks |= idx2binblock(ii))
1572 #define clear_binblock(ii) (binblocks &= ~(idx2binblock(ii)))
1578 /* Other static bookkeeping data */
1580 /* variables holding tunable values */
1582 static unsigned long trim_threshold = DEFAULT_TRIM_THRESHOLD;
1583 static unsigned long top_pad = DEFAULT_TOP_PAD;
1584 static unsigned int n_mmaps_max = DEFAULT_MMAP_MAX;
1585 static unsigned long mmap_threshold = DEFAULT_MMAP_THRESHOLD;
1587 /* The first value returned from sbrk */
1588 static char* sbrk_base = (char*)(-1);
1590 /* The maximum memory obtained from system via sbrk */
1591 static unsigned long max_sbrked_mem = 0;
1593 /* The maximum via either sbrk or mmap */
1594 static unsigned long max_total_mem = 0;
1596 /* internal working copy of mallinfo */
1597 static struct mallinfo current_mallinfo = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
1599 /* The total memory obtained from system via sbrk */
1600 #define sbrked_mem (current_mallinfo.arena)
1602 /* Tracking mmaps */
1605 static unsigned int n_mmaps = 0;
1607 static unsigned long mmapped_mem = 0;
1609 static unsigned int max_n_mmaps = 0;
1610 static unsigned long max_mmapped_mem = 0;
1623 These routines make a number of assertions about the states
1624 of data structures that should be true at all times. If any
1625 are not true, it's very likely that a user program has somehow
1626 trashed memory. (It's also possible that there is a coding error
1627 in malloc. In which case, please report it!)
1631 static void do_check_chunk(mchunkptr p)
1633 static void do_check_chunk(p) mchunkptr p;
1636 #if 0 /* causes warnings because assert() is off */
1637 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1640 /* No checkable chunk is mmapped */
1641 assert(!chunk_is_mmapped(p));
1643 /* Check for legal address ... */
1644 assert((char*)p >= sbrk_base);
1646 assert((char*)p + sz <= (char*)top);
1648 assert((char*)p + sz <= sbrk_base + sbrked_mem);
1654 static void do_check_free_chunk(mchunkptr p)
1656 static void do_check_free_chunk(p) mchunkptr p;
1659 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1660 #if 0 /* causes warnings because assert() is off */
1661 mchunkptr next = chunk_at_offset(p, sz);
1666 /* Check whether it claims to be free ... */
1669 /* Unless a special marker, must have OK fields */
1670 if ((long)sz >= (long)MINSIZE)
1672 assert((sz & MALLOC_ALIGN_MASK) == 0);
1673 assert(aligned_OK(chunk2mem(p)));
1674 /* ... matching footer field */
1675 assert(next->prev_size == sz);
1676 /* ... and is fully consolidated */
1677 assert(prev_inuse(p));
1678 assert (next == top || inuse(next));
1680 /* ... and has minimally sane links */
1681 assert(p->fd->bk == p);
1682 assert(p->bk->fd == p);
1684 else /* markers are always of size SIZE_SZ */
1685 assert(sz == SIZE_SZ);
1689 static void do_check_inuse_chunk(mchunkptr p)
1691 static void do_check_inuse_chunk(p) mchunkptr p;
1694 mchunkptr next = next_chunk(p);
1697 /* Check whether it claims to be in use ... */
1700 /* ... and is surrounded by OK chunks.
1701 Since more things can be checked with free chunks than inuse ones,
1702 if an inuse chunk borders them and debug is on, it's worth doing them.
1706 mchunkptr prv = prev_chunk(p);
1707 assert(next_chunk(prv) == p);
1708 do_check_free_chunk(prv);
1712 assert(prev_inuse(next));
1713 assert(chunksize(next) >= MINSIZE);
1715 else if (!inuse(next))
1716 do_check_free_chunk(next);
1721 static void do_check_malloced_chunk(mchunkptr p, INTERNAL_SIZE_T s)
1723 static void do_check_malloced_chunk(p, s) mchunkptr p; INTERNAL_SIZE_T s;
1726 #if 0 /* causes warnings because assert() is off */
1727 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1731 do_check_inuse_chunk(p);
1733 /* Legal size ... */
1734 assert((long)sz >= (long)MINSIZE);
1735 assert((sz & MALLOC_ALIGN_MASK) == 0);
1737 assert(room < (long)MINSIZE);
1739 /* ... and alignment */
1740 assert(aligned_OK(chunk2mem(p)));
1743 /* ... and was allocated at front of an available chunk */
1744 assert(prev_inuse(p));
1749 #define check_free_chunk(P) do_check_free_chunk(P)
1750 #define check_inuse_chunk(P) do_check_inuse_chunk(P)
1751 #define check_chunk(P) do_check_chunk(P)
1752 #define check_malloced_chunk(P,N) do_check_malloced_chunk(P,N)
1754 #define check_free_chunk(P)
1755 #define check_inuse_chunk(P)
1756 #define check_chunk(P)
1757 #define check_malloced_chunk(P,N)
1763 Macro-based internal utilities
1768 Linking chunks in bin lists.
1769 Call these only with variables, not arbitrary expressions, as arguments.
1773 Place chunk p of size s in its bin, in size order,
1774 putting it ahead of others of same size.
1778 #define frontlink(P, S, IDX, BK, FD) \
1780 if (S < MAX_SMALLBIN_SIZE) \
1782 IDX = smallbin_index(S); \
1783 mark_binblock(IDX); \
1788 FD->bk = BK->fd = P; \
1792 IDX = bin_index(S); \
1795 if (FD == BK) mark_binblock(IDX); \
1798 while (FD != BK && S < chunksize(FD)) FD = FD->fd; \
1803 FD->bk = BK->fd = P; \
1808 /* take a chunk off a list */
1810 #define unlink(P, BK, FD) \
1818 /* Place p as the last remainder */
1820 #define link_last_remainder(P) \
1822 last_remainder->fd = last_remainder->bk = P; \
1823 P->fd = P->bk = last_remainder; \
1826 /* Clear the last_remainder bin */
1828 #define clear_last_remainder \
1829 (last_remainder->fd = last_remainder->bk = last_remainder)
1836 /* Routines dealing with mmap(). */
1841 static mchunkptr mmap_chunk(size_t size)
1843 static mchunkptr mmap_chunk(size) size_t size;
1846 size_t page_mask = malloc_getpagesize - 1;
1849 #ifndef MAP_ANONYMOUS
1853 if(n_mmaps >= n_mmaps_max) return 0; /* too many regions */
1855 /* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
1856 * there is no following chunk whose prev_size field could be used.
1858 size = (size + SIZE_SZ + page_mask) & ~page_mask;
1860 #ifdef MAP_ANONYMOUS
1861 p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE,
1862 MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
1863 #else /* !MAP_ANONYMOUS */
1866 fd = open("/dev/zero", O_RDWR);
1867 if(fd < 0) return 0;
1869 p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
1872 if(p == (mchunkptr)-1) return 0;
1875 if (n_mmaps > max_n_mmaps) max_n_mmaps = n_mmaps;
1877 /* We demand that eight bytes into a page must be 8-byte aligned. */
1878 assert(aligned_OK(chunk2mem(p)));
1880 /* The offset to the start of the mmapped region is stored
1881 * in the prev_size field of the chunk; normally it is zero,
1882 * but that can be changed in memalign().
1885 set_head(p, size|IS_MMAPPED);
1887 mmapped_mem += size;
1888 if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1889 max_mmapped_mem = mmapped_mem;
1890 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1891 max_total_mem = mmapped_mem + sbrked_mem;
1896 static void munmap_chunk(mchunkptr p)
1898 static void munmap_chunk(p) mchunkptr p;
1901 INTERNAL_SIZE_T size = chunksize(p);
1904 assert (chunk_is_mmapped(p));
1905 assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1906 assert((n_mmaps > 0));
1907 assert(((p->prev_size + size) & (malloc_getpagesize-1)) == 0);
1910 mmapped_mem -= (size + p->prev_size);
1912 ret = munmap((char *)p - p->prev_size, size + p->prev_size);
1914 /* munmap returns non-zero on failure */
1921 static mchunkptr mremap_chunk(mchunkptr p, size_t new_size)
1923 static mchunkptr mremap_chunk(p, new_size) mchunkptr p; size_t new_size;
1926 size_t page_mask = malloc_getpagesize - 1;
1927 INTERNAL_SIZE_T offset = p->prev_size;
1928 INTERNAL_SIZE_T size = chunksize(p);
1931 assert (chunk_is_mmapped(p));
1932 assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1933 assert((n_mmaps > 0));
1934 assert(((size + offset) & (malloc_getpagesize-1)) == 0);
1936 /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
1937 new_size = (new_size + offset + SIZE_SZ + page_mask) & ~page_mask;
1939 cp = (char *)mremap((char *)p - offset, size + offset, new_size, 1);
1941 if (cp == (char *)-1) return 0;
1943 p = (mchunkptr)(cp + offset);
1945 assert(aligned_OK(chunk2mem(p)));
1947 assert((p->prev_size == offset));
1948 set_head(p, (new_size - offset)|IS_MMAPPED);
1950 mmapped_mem -= size + offset;
1951 mmapped_mem += new_size;
1952 if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1953 max_mmapped_mem = mmapped_mem;
1954 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1955 max_total_mem = mmapped_mem + sbrked_mem;
1959 #endif /* HAVE_MREMAP */
1961 #endif /* HAVE_MMAP */
1967 Extend the top-most chunk by obtaining memory from system.
1968 Main interface to sbrk (but see also malloc_trim).
1972 static void malloc_extend_top(INTERNAL_SIZE_T nb)
1974 static void malloc_extend_top(nb) INTERNAL_SIZE_T nb;
1977 char* brk; /* return value from sbrk */
1978 INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of sbrked space */
1979 INTERNAL_SIZE_T correction; /* bytes for 2nd sbrk call */
1980 char* new_brk; /* return of 2nd sbrk call */
1981 INTERNAL_SIZE_T top_size; /* new size of top chunk */
1983 mchunkptr old_top = top; /* Record state of old top */
1984 INTERNAL_SIZE_T old_top_size = chunksize(old_top);
1985 char* old_end = (char*)(chunk_at_offset(old_top, old_top_size));
1987 /* Pad request with top_pad plus minimal overhead */
1989 INTERNAL_SIZE_T sbrk_size = nb + top_pad + MINSIZE;
1990 unsigned long pagesz = malloc_getpagesize;
1992 /* If not the first time through, round to preserve page boundary */
1993 /* Otherwise, we need to correct to a page size below anyway. */
1994 /* (We also correct below if an intervening foreign sbrk call.) */
1996 if (sbrk_base != (char*)(-1))
1997 sbrk_size = (sbrk_size + (pagesz - 1)) & ~(pagesz - 1);
1999 brk = (char*)(MORECORE (sbrk_size));
2001 /* Fail if sbrk failed or if a foreign sbrk call killed our space */
2002 if (brk == (char*)(MORECORE_FAILURE) ||
2003 (brk < old_end && old_top != initial_top))
2006 sbrked_mem += sbrk_size;
2008 if (brk == old_end) /* can just add bytes to current top */
2010 top_size = sbrk_size + old_top_size;
2011 set_head(top, top_size | PREV_INUSE);
2015 if (sbrk_base == (char*)(-1)) /* First time through. Record base */
2017 else /* Someone else called sbrk(). Count those bytes as sbrked_mem. */
2018 sbrked_mem += brk - (char*)old_end;
2020 /* Guarantee alignment of first new chunk made from this space */
2021 front_misalign = (unsigned long)chunk2mem(brk) & MALLOC_ALIGN_MASK;
2022 if (front_misalign > 0)
2024 correction = (MALLOC_ALIGNMENT) - front_misalign;
2030 /* Guarantee the next brk will be at a page boundary */
2032 correction += ((((unsigned long)(brk + sbrk_size))+(pagesz-1)) &
2033 ~(pagesz - 1)) - ((unsigned long)(brk + sbrk_size));
2035 /* Allocate correction */
2036 new_brk = (char*)(MORECORE (correction));
2037 if (new_brk == (char*)(MORECORE_FAILURE)) return;
2039 sbrked_mem += correction;
2041 top = (mchunkptr)brk;
2042 top_size = new_brk - brk + correction;
2043 set_head(top, top_size | PREV_INUSE);
2045 if (old_top != initial_top)
2048 /* There must have been an intervening foreign sbrk call. */
2049 /* A double fencepost is necessary to prevent consolidation */
2051 /* If not enough space to do this, then user did something very wrong */
2052 if (old_top_size < MINSIZE)
2054 set_head(top, PREV_INUSE); /* will force null return from malloc */
2058 /* Also keep size a multiple of MALLOC_ALIGNMENT */
2059 old_top_size = (old_top_size - 3*SIZE_SZ) & ~MALLOC_ALIGN_MASK;
2060 set_head_size(old_top, old_top_size);
2061 chunk_at_offset(old_top, old_top_size )->size =
2063 chunk_at_offset(old_top, old_top_size + SIZE_SZ)->size =
2065 /* If possible, release the rest. */
2066 if (old_top_size >= MINSIZE)
2067 fREe(chunk2mem(old_top));
2071 if ((unsigned long)sbrked_mem > (unsigned long)max_sbrked_mem)
2072 max_sbrked_mem = sbrked_mem;
2073 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
2074 max_total_mem = mmapped_mem + sbrked_mem;
2076 /* We always land on a page boundary */
2077 assert(((unsigned long)((char*)top + top_size) & (pagesz - 1)) == 0);
2083 /* Main public routines */
2089 The requested size is first converted into a usable form, `nb'.
2090 This currently means to add 4 bytes overhead plus possibly more to
2091 obtain 8-byte alignment and/or to obtain a size of at least
2092 MINSIZE (currently 16 bytes), the smallest allocatable size.
2093 (All fits are considered `exact' if they are within MINSIZE bytes.)
2095 From there, the first successful of the following steps is taken:
2097 1. The bin corresponding to the request size is scanned, and if
2098 a chunk of exactly the right size is found, it is taken.
2100 2. The most recently remaindered chunk is used if it is big
2101 enough. This is a form of (roving) first fit, used only in
2102 the absence of exact fits. Runs of consecutive requests use
2103 the remainder of the chunk used for the previous such request
2104 whenever possible. This limited use of a first-fit style
2105 allocation strategy tends to give contiguous chunks
2106 coextensive lifetimes, which improves locality and can reduce
2107 fragmentation in the long run.
2109 3. Other bins are scanned in increasing size order, using a
2110 chunk big enough to fulfill the request, and splitting off
2111 any remainder. This search is strictly by best-fit; i.e.,
2112 the smallest (with ties going to approximately the least
2113 recently used) chunk that fits is selected.
2115 4. If large enough, the chunk bordering the end of memory
2116 (`top') is split off. (This use of `top' is in accord with
2117 the best-fit search rule. In effect, `top' is treated as
2118 larger (and thus less well fitting) than any other available
2119 chunk since it can be extended to be as large as necessary
2120 (up to system limitations).
2122 5. If the request size meets the mmap threshold and the
2123 system supports mmap, and there are few enough currently
2124 allocated mmapped regions, and a call to mmap succeeds,
2125 the request is allocated via direct memory mapping.
2127 6. Otherwise, the top of memory is extended by
2128 obtaining more space from the system (normally using sbrk,
2129 but definable to anything else via the MORECORE macro).
2130 Memory is gathered from the system (in system page-sized
2131 units) in a way that allows chunks obtained across different
2132 sbrk calls to be consolidated, but does not require
2133 contiguous memory. Thus, it should be safe to intersperse
2134 mallocs with other sbrk calls.
2137 All allocations are made from the the `lowest' part of any found
2138 chunk. (The implementation invariant is that prev_inuse is
2139 always true of any allocated chunk; i.e., that each allocated
2140 chunk borders either a previously allocated and still in-use chunk,
2141 or the base of its memory arena.)
2146 Void_t* mALLOc(size_t bytes)
2148 Void_t* mALLOc(bytes) size_t bytes;
2151 mchunkptr victim; /* inspected/selected chunk */
2152 INTERNAL_SIZE_T victim_size; /* its size */
2153 int idx; /* index for bin traversal */
2154 mbinptr bin; /* associated bin */
2155 mchunkptr remainder; /* remainder from a split */
2156 long remainder_size; /* its size */
2157 int remainder_index; /* its bin index */
2158 unsigned long block; /* block traverser bit */
2159 int startidx; /* first bin of a traversed block */
2160 mchunkptr fwd; /* misc temp for linking */
2161 mchunkptr bck; /* misc temp for linking */
2162 mbinptr q; /* misc temp */
2166 if ((long)bytes < 0) return 0;
2168 nb = request2size(bytes); /* padded request size; */
2170 /* Check for exact match in a bin */
2172 if (is_small_request(nb)) /* Faster version for small requests */
2174 idx = smallbin_index(nb);
2176 /* No traversal or size check necessary for small bins. */
2181 /* Also scan the next one, since it would have a remainder < MINSIZE */
2189 victim_size = chunksize(victim);
2190 unlink(victim, bck, fwd);
2191 set_inuse_bit_at_offset(victim, victim_size);
2192 check_malloced_chunk(victim, nb);
2193 return chunk2mem(victim);
2196 idx += 2; /* Set for bin scan below. We've already scanned 2 bins. */
2201 idx = bin_index(nb);
2204 for (victim = last(bin); victim != bin; victim = victim->bk)
2206 victim_size = chunksize(victim);
2207 remainder_size = victim_size - nb;
2209 if (remainder_size >= (long)MINSIZE) /* too big */
2211 --idx; /* adjust to rescan below after checking last remainder */
2215 else if (remainder_size >= 0) /* exact fit */
2217 unlink(victim, bck, fwd);
2218 set_inuse_bit_at_offset(victim, victim_size);
2219 check_malloced_chunk(victim, nb);
2220 return chunk2mem(victim);
2228 /* Try to use the last split-off remainder */
2230 if ( (victim = last_remainder->fd) != last_remainder)
2232 victim_size = chunksize(victim);
2233 remainder_size = victim_size - nb;
2235 if (remainder_size >= (long)MINSIZE) /* re-split */
2237 remainder = chunk_at_offset(victim, nb);
2238 set_head(victim, nb | PREV_INUSE);
2239 link_last_remainder(remainder);
2240 set_head(remainder, remainder_size | PREV_INUSE);
2241 set_foot(remainder, remainder_size);
2242 check_malloced_chunk(victim, nb);
2243 return chunk2mem(victim);
2246 clear_last_remainder;
2248 if (remainder_size >= 0) /* exhaust */
2250 set_inuse_bit_at_offset(victim, victim_size);
2251 check_malloced_chunk(victim, nb);
2252 return chunk2mem(victim);
2255 /* Else place in bin */
2257 frontlink(victim, victim_size, remainder_index, bck, fwd);
2261 If there are any possibly nonempty big-enough blocks,
2262 search for best fitting chunk by scanning bins in blockwidth units.
2265 if ( (block = idx2binblock(idx)) <= binblocks)
2268 /* Get to the first marked block */
2270 if ( (block & binblocks) == 0)
2272 /* force to an even block boundary */
2273 idx = (idx & ~(BINBLOCKWIDTH - 1)) + BINBLOCKWIDTH;
2275 while ((block & binblocks) == 0)
2277 idx += BINBLOCKWIDTH;
2282 /* For each possibly nonempty block ... */
2285 startidx = idx; /* (track incomplete blocks) */
2286 q = bin = bin_at(idx);
2288 /* For each bin in this block ... */
2291 /* Find and use first big enough chunk ... */
2293 for (victim = last(bin); victim != bin; victim = victim->bk)
2295 victim_size = chunksize(victim);
2296 remainder_size = victim_size - nb;
2298 if (remainder_size >= (long)MINSIZE) /* split */
2300 remainder = chunk_at_offset(victim, nb);
2301 set_head(victim, nb | PREV_INUSE);
2302 unlink(victim, bck, fwd);
2303 link_last_remainder(remainder);
2304 set_head(remainder, remainder_size | PREV_INUSE);
2305 set_foot(remainder, remainder_size);
2306 check_malloced_chunk(victim, nb);
2307 return chunk2mem(victim);
2310 else if (remainder_size >= 0) /* take */
2312 set_inuse_bit_at_offset(victim, victim_size);
2313 unlink(victim, bck, fwd);
2314 check_malloced_chunk(victim, nb);
2315 return chunk2mem(victim);
2320 bin = next_bin(bin);
2322 } while ((++idx & (BINBLOCKWIDTH - 1)) != 0);
2324 /* Clear out the block bit. */
2326 do /* Possibly backtrack to try to clear a partial block */
2328 if ((startidx & (BINBLOCKWIDTH - 1)) == 0)
2330 binblocks &= ~block;
2335 } while (first(q) == q);
2337 /* Get to the next possibly nonempty block */
2339 if ( (block <<= 1) <= binblocks && (block != 0) )
2341 while ((block & binblocks) == 0)
2343 idx += BINBLOCKWIDTH;
2353 /* Try to use top chunk */
2355 /* Require that there be a remainder, ensuring top always exists */
2356 if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2360 /* If big and would otherwise need to extend, try to use mmap instead */
2361 if ((unsigned long)nb >= (unsigned long)mmap_threshold &&
2362 (victim = mmap_chunk(nb)) != 0)
2363 return chunk2mem(victim);
2367 malloc_extend_top(nb);
2368 if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2369 return 0; /* propagate failure */
2373 set_head(victim, nb | PREV_INUSE);
2374 top = chunk_at_offset(victim, nb);
2375 set_head(top, remainder_size | PREV_INUSE);
2376 check_malloced_chunk(victim, nb);
2377 return chunk2mem(victim);
2390 1. free(0) has no effect.
2392 2. If the chunk was allocated via mmap, it is release via munmap().
2394 3. If a returned chunk borders the current high end of memory,
2395 it is consolidated into the top, and if the total unused
2396 topmost memory exceeds the trim threshold, malloc_trim is
2399 4. Other chunks are consolidated as they arrive, and
2400 placed in corresponding bins. (This includes the case of
2401 consolidating with the current `last_remainder').
2407 void fREe(Void_t* mem)
2409 void fREe(mem) Void_t* mem;
2412 mchunkptr p; /* chunk corresponding to mem */
2413 INTERNAL_SIZE_T hd; /* its head field */
2414 INTERNAL_SIZE_T sz; /* its size */
2415 int idx; /* its bin index */
2416 mchunkptr next; /* next contiguous chunk */
2417 INTERNAL_SIZE_T nextsz; /* its size */
2418 INTERNAL_SIZE_T prevsz; /* size of previous contiguous chunk */
2419 mchunkptr bck; /* misc temp for linking */
2420 mchunkptr fwd; /* misc temp for linking */
2421 int islr; /* track whether merging with last_remainder */
2423 if (mem == 0) /* free(0) has no effect */
2430 if (hd & IS_MMAPPED) /* release mmapped memory. */
2437 check_inuse_chunk(p);
2439 sz = hd & ~PREV_INUSE;
2440 next = chunk_at_offset(p, sz);
2441 nextsz = chunksize(next);
2443 if (next == top) /* merge with top */
2447 if (!(hd & PREV_INUSE)) /* consolidate backward */
2449 prevsz = p->prev_size;
2450 p = chunk_at_offset(p, -((long) prevsz));
2452 unlink(p, bck, fwd);
2455 set_head(p, sz | PREV_INUSE);
2457 if ((unsigned long)(sz) >= (unsigned long)trim_threshold)
2458 malloc_trim(top_pad);
2462 set_head(next, nextsz); /* clear inuse bit */
2466 if (!(hd & PREV_INUSE)) /* consolidate backward */
2468 prevsz = p->prev_size;
2469 p = chunk_at_offset(p, -((long) prevsz));
2472 if (p->fd == last_remainder) /* keep as last_remainder */
2475 unlink(p, bck, fwd);
2478 if (!(inuse_bit_at_offset(next, nextsz))) /* consolidate forward */
2482 if (!islr && next->fd == last_remainder) /* re-insert last_remainder */
2485 link_last_remainder(p);
2488 unlink(next, bck, fwd);
2492 set_head(p, sz | PREV_INUSE);
2495 frontlink(p, sz, idx, bck, fwd);
2506 Chunks that were obtained via mmap cannot be extended or shrunk
2507 unless HAVE_MREMAP is defined, in which case mremap is used.
2508 Otherwise, if their reallocation is for additional space, they are
2509 copied. If for less, they are just left alone.
2511 Otherwise, if the reallocation is for additional space, and the
2512 chunk can be extended, it is, else a malloc-copy-free sequence is
2513 taken. There are several different ways that a chunk could be
2514 extended. All are tried:
2516 * Extending forward into following adjacent free chunk.
2517 * Shifting backwards, joining preceding adjacent space
2518 * Both shifting backwards and extending forward.
2519 * Extending into newly sbrked space
2521 Unless the #define REALLOC_ZERO_BYTES_FREES is set, realloc with a
2522 size argument of zero (re)allocates a minimum-sized chunk.
2524 If the reallocation is for less space, and the new request is for
2525 a `small' (<512 bytes) size, then the newly unused space is lopped
2528 The old unix realloc convention of allowing the last-free'd chunk
2529 to be used as an argument to realloc is no longer supported.
2530 I don't know of any programs still relying on this feature,
2531 and allowing it would also allow too many other incorrect
2532 usages of realloc to be sensible.
2539 Void_t* rEALLOc(Void_t* oldmem, size_t bytes)
2541 Void_t* rEALLOc(oldmem, bytes) Void_t* oldmem; size_t bytes;
2544 INTERNAL_SIZE_T nb; /* padded request size */
2546 mchunkptr oldp; /* chunk corresponding to oldmem */
2547 INTERNAL_SIZE_T oldsize; /* its size */
2549 mchunkptr newp; /* chunk to return */
2550 INTERNAL_SIZE_T newsize; /* its size */
2551 Void_t* newmem; /* corresponding user mem */
2553 mchunkptr next; /* next contiguous chunk after oldp */
2554 INTERNAL_SIZE_T nextsize; /* its size */
2556 mchunkptr prev; /* previous contiguous chunk before oldp */
2557 INTERNAL_SIZE_T prevsize; /* its size */
2559 mchunkptr remainder; /* holds split off extra space from newp */
2560 INTERNAL_SIZE_T remainder_size; /* its size */
2562 mchunkptr bck; /* misc temp for linking */
2563 mchunkptr fwd; /* misc temp for linking */
2565 #ifdef REALLOC_ZERO_BYTES_FREES
2566 if (bytes == 0) { fREe(oldmem); return 0; }
2569 if ((long)bytes < 0) return 0;
2571 /* realloc of null is supposed to be same as malloc */
2572 if (oldmem == 0) return mALLOc(bytes);
2574 newp = oldp = mem2chunk(oldmem);
2575 newsize = oldsize = chunksize(oldp);
2578 nb = request2size(bytes);
2581 if (chunk_is_mmapped(oldp))
2584 newp = mremap_chunk(oldp, nb);
2585 if(newp) return chunk2mem(newp);
2587 /* Note the extra SIZE_SZ overhead. */
2588 if(oldsize - SIZE_SZ >= nb) return oldmem; /* do nothing */
2589 /* Must alloc, copy, free. */
2590 newmem = mALLOc(bytes);
2591 if (newmem == 0) return 0; /* propagate failure */
2592 MALLOC_COPY(newmem, oldmem, oldsize - 2*SIZE_SZ);
2598 check_inuse_chunk(oldp);
2600 if ((long)(oldsize) < (long)(nb))
2603 /* Try expanding forward */
2605 next = chunk_at_offset(oldp, oldsize);
2606 if (next == top || !inuse(next))
2608 nextsize = chunksize(next);
2610 /* Forward into top only if a remainder */
2613 if ((long)(nextsize + newsize) >= (long)(nb + MINSIZE))
2615 newsize += nextsize;
2616 top = chunk_at_offset(oldp, nb);
2617 set_head(top, (newsize - nb) | PREV_INUSE);
2618 set_head_size(oldp, nb);
2619 return chunk2mem(oldp);
2623 /* Forward into next chunk */
2624 else if (((long)(nextsize + newsize) >= (long)(nb)))
2626 unlink(next, bck, fwd);
2627 newsize += nextsize;
2637 /* Try shifting backwards. */
2639 if (!prev_inuse(oldp))
2641 prev = prev_chunk(oldp);
2642 prevsize = chunksize(prev);
2644 /* try forward + backward first to save a later consolidation */
2651 if ((long)(nextsize + prevsize + newsize) >= (long)(nb + MINSIZE))
2653 unlink(prev, bck, fwd);
2655 newsize += prevsize + nextsize;
2656 newmem = chunk2mem(newp);
2657 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2658 top = chunk_at_offset(newp, nb);
2659 set_head(top, (newsize - nb) | PREV_INUSE);
2660 set_head_size(newp, nb);
2665 /* into next chunk */
2666 else if (((long)(nextsize + prevsize + newsize) >= (long)(nb)))
2668 unlink(next, bck, fwd);
2669 unlink(prev, bck, fwd);
2671 newsize += nextsize + prevsize;
2672 newmem = chunk2mem(newp);
2673 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2679 if (prev != 0 && (long)(prevsize + newsize) >= (long)nb)
2681 unlink(prev, bck, fwd);
2683 newsize += prevsize;
2684 newmem = chunk2mem(newp);
2685 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2692 newmem = mALLOc (bytes);
2694 if (newmem == 0) /* propagate failure */
2697 /* Avoid copy if newp is next chunk after oldp. */
2698 /* (This can only happen when new chunk is sbrk'ed.) */
2700 if ( (newp = mem2chunk(newmem)) == next_chunk(oldp))
2702 newsize += chunksize(newp);
2707 /* Otherwise copy, free, and exit */
2708 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2714 split: /* split off extra room in old or expanded chunk */
2716 if (newsize - nb >= MINSIZE) /* split off remainder */
2718 remainder = chunk_at_offset(newp, nb);
2719 remainder_size = newsize - nb;
2720 set_head_size(newp, nb);
2721 set_head(remainder, remainder_size | PREV_INUSE);
2722 set_inuse_bit_at_offset(remainder, remainder_size);
2723 fREe(chunk2mem(remainder)); /* let free() deal with it */
2727 set_head_size(newp, newsize);
2728 set_inuse_bit_at_offset(newp, newsize);
2731 check_inuse_chunk(newp);
2732 return chunk2mem(newp);
2742 memalign requests more than enough space from malloc, finds a spot
2743 within that chunk that meets the alignment request, and then
2744 possibly frees the leading and trailing space.
2746 The alignment argument must be a power of two. This property is not
2747 checked by memalign, so misuse may result in random runtime errors.
2749 8-byte alignment is guaranteed by normal malloc calls, so don't
2750 bother calling memalign with an argument of 8 or less.
2752 Overreliance on memalign is a sure way to fragment space.
2758 Void_t* mEMALIGn(size_t alignment, size_t bytes)
2760 Void_t* mEMALIGn(alignment, bytes) size_t alignment; size_t bytes;
2763 INTERNAL_SIZE_T nb; /* padded request size */
2764 char* m; /* memory returned by malloc call */
2765 mchunkptr p; /* corresponding chunk */
2766 char* brk; /* alignment point within p */
2767 mchunkptr newp; /* chunk to return */
2768 INTERNAL_SIZE_T newsize; /* its size */
2769 INTERNAL_SIZE_T leadsize; /* leading space befor alignment point */
2770 mchunkptr remainder; /* spare room at end to split off */
2771 long remainder_size; /* its size */
2773 if ((long)bytes < 0) return 0;
2775 /* If need less alignment than we give anyway, just relay to malloc */
2777 if (alignment <= MALLOC_ALIGNMENT) return mALLOc(bytes);
2779 /* Otherwise, ensure that it is at least a minimum chunk size */
2781 if (alignment < MINSIZE) alignment = MINSIZE;
2783 /* Call malloc with worst case padding to hit alignment. */
2785 nb = request2size(bytes);
2786 m = (char*)(mALLOc(nb + alignment + MINSIZE));
2788 if (m == 0) return 0; /* propagate failure */
2792 if ((((unsigned long)(m)) % alignment) == 0) /* aligned */
2795 if(chunk_is_mmapped(p))
2796 return chunk2mem(p); /* nothing more to do */
2799 else /* misaligned */
2802 Find an aligned spot inside chunk.
2803 Since we need to give back leading space in a chunk of at
2804 least MINSIZE, if the first calculation places us at
2805 a spot with less than MINSIZE leader, we can move to the
2806 next aligned spot -- we've allocated enough total room so that
2807 this is always possible.
2810 brk = (char*)mem2chunk(((unsigned long)(m + alignment - 1)) & -((signed) alignment));
2811 if ((long)(brk - (char*)(p)) < MINSIZE) brk = brk + alignment;
2813 newp = (mchunkptr)brk;
2814 leadsize = brk - (char*)(p);
2815 newsize = chunksize(p) - leadsize;
2818 if(chunk_is_mmapped(p))
2820 newp->prev_size = p->prev_size + leadsize;
2821 set_head(newp, newsize|IS_MMAPPED);
2822 return chunk2mem(newp);
2826 /* give back leader, use the rest */
2828 set_head(newp, newsize | PREV_INUSE);
2829 set_inuse_bit_at_offset(newp, newsize);
2830 set_head_size(p, leadsize);
2834 assert (newsize >= nb && (((unsigned long)(chunk2mem(p))) % alignment) == 0);
2837 /* Also give back spare room at the end */
2839 remainder_size = chunksize(p) - nb;
2841 if (remainder_size >= (long)MINSIZE)
2843 remainder = chunk_at_offset(p, nb);
2844 set_head(remainder, remainder_size | PREV_INUSE);
2845 set_head_size(p, nb);
2846 fREe(chunk2mem(remainder));
2849 check_inuse_chunk(p);
2850 return chunk2mem(p);
2858 valloc just invokes memalign with alignment argument equal
2859 to the page size of the system (or as near to this as can
2860 be figured out from all the includes/defines above.)
2864 Void_t* vALLOc(size_t bytes)
2866 Void_t* vALLOc(bytes) size_t bytes;
2869 return mEMALIGn (malloc_getpagesize, bytes);
2873 pvalloc just invokes valloc for the nearest pagesize
2874 that will accommodate request
2879 Void_t* pvALLOc(size_t bytes)
2881 Void_t* pvALLOc(bytes) size_t bytes;
2884 size_t pagesize = malloc_getpagesize;
2885 return mEMALIGn (pagesize, (bytes + pagesize - 1) & ~(pagesize - 1));
2890 calloc calls malloc, then zeroes out the allocated chunk.
2895 Void_t* cALLOc(size_t n, size_t elem_size)
2897 Void_t* cALLOc(n, elem_size) size_t n; size_t elem_size;
2901 INTERNAL_SIZE_T csz;
2903 INTERNAL_SIZE_T sz = n * elem_size;
2906 /* check if expand_top called, in which case don't need to clear */
2908 mchunkptr oldtop = top;
2909 INTERNAL_SIZE_T oldtopsize = chunksize(top);
2911 Void_t* mem = mALLOc (sz);
2913 if ((long)n < 0) return 0;
2921 /* Two optional cases in which clearing not necessary */
2925 if (chunk_is_mmapped(p)) return mem;
2931 if (p == oldtop && csz > oldtopsize)
2933 /* clear only the bytes from non-freshly-sbrked memory */
2938 MALLOC_ZERO(mem, csz - SIZE_SZ);
2945 cfree just calls free. It is needed/defined on some systems
2946 that pair it with calloc, presumably for odd historical reasons.
2950 #if !defined(INTERNAL_LINUX_C_LIB) || !defined(__ELF__)
2952 void cfree(Void_t *mem)
2954 void cfree(mem) Void_t *mem;
2965 Malloc_trim gives memory back to the system (via negative
2966 arguments to sbrk) if there is unused memory at the `high' end of
2967 the malloc pool. You can call this after freeing large blocks of
2968 memory to potentially reduce the system-level memory requirements
2969 of a program. However, it cannot guarantee to reduce memory. Under
2970 some allocation patterns, some large free blocks of memory will be
2971 locked between two used chunks, so they cannot be given back to
2974 The `pad' argument to malloc_trim represents the amount of free
2975 trailing space to leave untrimmed. If this argument is zero,
2976 only the minimum amount of memory to maintain internal data
2977 structures will be left (one page or less). Non-zero arguments
2978 can be supplied to maintain enough trailing space to service
2979 future expected allocations without having to re-obtain memory
2982 Malloc_trim returns 1 if it actually released any memory, else 0.
2987 int malloc_trim(size_t pad)
2989 int malloc_trim(pad) size_t pad;
2992 long top_size; /* Amount of top-most memory */
2993 long extra; /* Amount to release */
2994 char* current_brk; /* address returned by pre-check sbrk call */
2995 char* new_brk; /* address returned by negative sbrk call */
2997 unsigned long pagesz = malloc_getpagesize;
2999 top_size = chunksize(top);
3000 extra = ((top_size - pad - MINSIZE + (pagesz-1)) / pagesz - 1) * pagesz;
3002 if (extra < (long)pagesz) /* Not enough memory to release */
3007 /* Test to make sure no one else called sbrk */
3008 current_brk = (char*)(MORECORE (0));
3009 if (current_brk != (char*)(top) + top_size)
3010 return 0; /* Apparently we don't own memory; must fail */
3014 new_brk = (char*)(MORECORE (-extra));
3016 if (new_brk == (char*)(MORECORE_FAILURE)) /* sbrk failed? */
3018 /* Try to figure out what we have */
3019 current_brk = (char*)(MORECORE (0));
3020 top_size = current_brk - (char*)top;
3021 if (top_size >= (long)MINSIZE) /* if not, we are very very dead! */
3023 sbrked_mem = current_brk - sbrk_base;
3024 set_head(top, top_size | PREV_INUSE);
3032 /* Success. Adjust top accordingly. */
3033 set_head(top, (top_size - extra) | PREV_INUSE);
3034 sbrked_mem -= extra;
3047 This routine tells you how many bytes you can actually use in an
3048 allocated chunk, which may be more than you requested (although
3049 often not). You can use this many bytes without worrying about
3050 overwriting other allocated objects. Not a particularly great
3051 programming practice, but still sometimes useful.
3056 size_t malloc_usable_size(Void_t* mem)
3058 size_t malloc_usable_size(mem) Void_t* mem;
3067 if(!chunk_is_mmapped(p))
3069 if (!inuse(p)) return 0;
3070 check_inuse_chunk(p);
3071 return chunksize(p) - SIZE_SZ;
3073 return chunksize(p) - 2*SIZE_SZ;
3080 /* Utility to update current_mallinfo for malloc_stats and mallinfo() */
3083 static void malloc_update_mallinfo()
3092 INTERNAL_SIZE_T avail = chunksize(top);
3093 int navail = ((long)(avail) >= (long)MINSIZE)? 1 : 0;
3095 for (i = 1; i < NAV; ++i)
3098 for (p = last(b); p != b; p = p->bk)
3101 check_free_chunk(p);
3102 for (q = next_chunk(p);
3103 q < top && inuse(q) && (long)(chunksize(q)) >= (long)MINSIZE;
3105 check_inuse_chunk(q);
3107 avail += chunksize(p);
3112 current_mallinfo.ordblks = navail;
3113 current_mallinfo.uordblks = sbrked_mem - avail;
3114 current_mallinfo.fordblks = avail;
3115 current_mallinfo.hblks = n_mmaps;
3116 current_mallinfo.hblkhd = mmapped_mem;
3117 current_mallinfo.keepcost = chunksize(top);
3128 Prints on the amount of space obtain from the system (both
3129 via sbrk and mmap), the maximum amount (which may be more than
3130 current if malloc_trim and/or munmap got called), the maximum
3131 number of simultaneous mmap regions used, and the current number
3132 of bytes allocated via malloc (or realloc, etc) but not yet
3133 freed. (Note that this is the number of bytes allocated, not the
3134 number requested. It will be larger than the number requested
3135 because of alignment and bookkeeping overhead.)
3142 malloc_update_mallinfo();
3143 printf("max system bytes = %10u\n",
3144 (unsigned int)(max_total_mem));
3145 printf("system bytes = %10u\n",
3146 (unsigned int)(sbrked_mem + mmapped_mem));
3147 printf("in use bytes = %10u\n",
3148 (unsigned int)(current_mallinfo.uordblks + mmapped_mem));
3150 printf("max mmap regions = %10u\n",
3151 (unsigned int)max_n_mmaps);
3157 mallinfo returns a copy of updated current mallinfo.
3161 struct mallinfo mALLINFo()
3163 malloc_update_mallinfo();
3164 return current_mallinfo;
3174 mallopt is the general SVID/XPG interface to tunable parameters.
3175 The format is to provide a (parameter-number, parameter-value) pair.
3176 mallopt then sets the corresponding parameter to the argument
3177 value if it can (i.e., so long as the value is meaningful),
3178 and returns 1 if successful else 0.
3180 See descriptions of tunable parameters above.
3185 int mALLOPt(int param_number, int value)
3187 int mALLOPt(param_number, value) int param_number; int value;
3190 switch(param_number)
3192 case M_TRIM_THRESHOLD:
3193 trim_threshold = value; return 1;
3195 top_pad = value; return 1;
3196 case M_MMAP_THRESHOLD:
3197 mmap_threshold = value; return 1;
3200 n_mmaps_max = value; return 1;
3202 if (value != 0) return 0; else n_mmaps_max = value; return 1;
3214 V2.6.6 Sun Dec 5 07:42:19 1999 Doug Lea (dl at gee)
3215 * return null for negative arguments
3217 * Add 'LACKS_SYS_PARAM_H' for those systems without 'sys/param.h'
3218 (e.g. WIN32 platforms)
3219 * Cleanup up header file inclusion for WIN32 platforms
3220 * Cleanup code to avoid Microsoft Visual C++ compiler complaints
3221 * Add 'USE_DL_PREFIX' to quickly allow co-existence with existing
3222 memory allocation routines
3223 * Set 'malloc_getpagesize' for WIN32 platforms (needs more work)
3224 * Use 'assert' rather than 'ASSERT' in WIN32 code to conform to
3225 usage of 'assert' in non-WIN32 code
3226 * Improve WIN32 'sbrk()' emulation's 'findRegion()' routine to
3228 * Always call 'fREe()' rather than 'free()'
3230 V2.6.5 Wed Jun 17 15:57:31 1998 Doug Lea (dl at gee)
3231 * Fixed ordering problem with boundary-stamping
3233 V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee)
3234 * Added pvalloc, as recommended by H.J. Liu
3235 * Added 64bit pointer support mainly from Wolfram Gloger
3236 * Added anonymously donated WIN32 sbrk emulation
3237 * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen
3238 * malloc_extend_top: fix mask error that caused wastage after
3240 * Add linux mremap support code from HJ Liu
3242 V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee)
3243 * Integrated most documentation with the code.
3244 * Add support for mmap, with help from
3246 * Use last_remainder in more cases.
3248 * Use ordered bins instead of best-fit threshhold
3249 * Eliminate block-local decls to simplify tracing and debugging.
3250 * Support another case of realloc via move into top
3251 * Fix error occuring when initial sbrk_base not word-aligned.
3252 * Rely on page size for units instead of SBRK_UNIT to
3253 avoid surprises about sbrk alignment conventions.
3254 * Add mallinfo, mallopt. Thanks to Raymond Nijssen
3256 * Add `pad' argument to malloc_trim and top_pad mallopt parameter.
3257 * More precautions for cases where other routines call sbrk,
3259 * Added macros etc., allowing use in linux libc from
3261 * Inverted this history list
3263 V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee)
3264 * Re-tuned and fixed to behave more nicely with V2.6.0 changes.
3265 * Removed all preallocation code since under current scheme
3266 the work required to undo bad preallocations exceeds
3267 the work saved in good cases for most test programs.
3268 * No longer use return list or unconsolidated bins since
3269 no scheme using them consistently outperforms those that don't
3270 given above changes.
3271 * Use best fit for very large chunks to prevent some worst-cases.
3272 * Added some support for debugging
3274 V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee)
3275 * Removed footers when chunks are in use. Thanks to
3278 V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee)
3279 * Added malloc_trim, with help from Wolfram Gloger
3282 V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g)
3284 V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g)
3285 * realloc: try to expand in both directions
3286 * malloc: swap order of clean-bin strategy;
3287 * realloc: only conditionally expand backwards
3288 * Try not to scavenge used bins
3289 * Use bin counts as a guide to preallocation
3290 * Occasionally bin return list chunks in first scan
3293 V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g)
3294 * faster bin computation & slightly different binning
3295 * merged all consolidations to one part of malloc proper
3296 (eliminating old malloc_find_space & malloc_clean_bin)
3297 * Scan 2 returns chunks (not just 1)
3298 * Propagate failure in realloc if malloc returns 0
3299 * Add stuff to allow compilation on non-ANSI compilers
3302 V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu)
3303 * removed potential for odd address access in prev_chunk
3304 * removed dependency on getpagesize.h
3305 * misc cosmetics and a bit more internal documentation
3306 * anticosmetics: mangled names in macros to evade debugger strangeness
3307 * tested on sparc, hp-700, dec-mips, rs6000
3308 with gcc & native cc (hp, dec only) allowing
3309 Detlefs & Zorn comparison study (in SIGPLAN Notices.)
3311 Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu)
3312 * Based loosely on libg++-1.2X malloc. (It retains some of the overall
3313 structure of old version, but most details differ.)