1 /* ---------- To make a malloc.h, start cutting here ------------ */
4 A version of malloc/free/realloc written by Doug Lea and released to the
5 public domain. Send questions/comments/complaints/performance data
8 * VERSION 2.6.6 Sun Mar 5 19:10:03 2000 Doug Lea (dl at gee)
10 Note: There may be an updated version of this malloc obtainable at
11 ftp://g.oswego.edu/pub/misc/malloc.c
12 Check before installing!
14 * Why use this malloc?
16 This is not the fastest, most space-conserving, most portable, or
17 most tunable malloc ever written. However it is among the fastest
18 while also being among the most space-conserving, portable and tunable.
19 Consistent balance across these factors results in a good general-purpose
20 allocator. For a high-level description, see
21 http://g.oswego.edu/dl/html/malloc.html
23 * Synopsis of public routines
25 (Much fuller descriptions are contained in the program documentation below.)
28 Return a pointer to a newly allocated chunk of at least n bytes, or null
29 if no space is available.
31 Release the chunk of memory pointed to by p, or no effect if p is null.
32 realloc(Void_t* p, size_t n);
33 Return a pointer to a chunk of size n that contains the same data
34 as does chunk p up to the minimum of (n, p's size) bytes, or null
35 if no space is available. The returned pointer may or may not be
36 the same as p. If p is null, equivalent to malloc. Unless the
37 #define REALLOC_ZERO_BYTES_FREES below is set, realloc with a
38 size argument of zero (re)allocates a minimum-sized chunk.
39 memalign(size_t alignment, size_t n);
40 Return a pointer to a newly allocated chunk of n bytes, aligned
41 in accord with the alignment argument, which must be a power of
44 Equivalent to memalign(pagesize, n), where pagesize is the page
45 size of the system (or as near to this as can be figured out from
46 all the includes/defines below.)
48 Equivalent to valloc(minimum-page-that-holds(n)), that is,
49 round up n to nearest pagesize.
50 calloc(size_t unit, size_t quantity);
51 Returns a pointer to quantity * unit bytes, with all locations
54 Equivalent to free(p).
55 malloc_trim(size_t pad);
56 Release all but pad bytes of freed top-most memory back
57 to the system. Return 1 if successful, else 0.
58 malloc_usable_size(Void_t* p);
59 Report the number usable allocated bytes associated with allocated
60 chunk p. This may or may not report more bytes than were requested,
61 due to alignment and minimum size constraints.
63 Prints brief summary statistics on stderr.
65 Returns (by copy) a struct containing various summary statistics.
66 mallopt(int parameter_number, int parameter_value)
67 Changes one of the tunable parameters described below. Returns
68 1 if successful in changing the parameter, else 0.
73 8 byte alignment is currently hardwired into the design. This
74 seems to suffice for all current machines and C compilers.
76 Assumed pointer representation: 4 or 8 bytes
77 Code for 8-byte pointers is untested by me but has worked
78 reliably by Wolfram Gloger, who contributed most of the
79 changes supporting this.
81 Assumed size_t representation: 4 or 8 bytes
82 Note that size_t is allowed to be 4 bytes even if pointers are 8.
84 Minimum overhead per allocated chunk: 4 or 8 bytes
85 Each malloced chunk has a hidden overhead of 4 bytes holding size
86 and status information.
88 Minimum allocated size: 4-byte ptrs: 16 bytes (including 4 overhead)
89 8-byte ptrs: 24/32 bytes (including, 4/8 overhead)
91 When a chunk is freed, 12 (for 4byte ptrs) or 20 (for 8 byte
92 ptrs but 4 byte size) or 24 (for 8/8) additional bytes are
93 needed; 4 (8) for a trailing size field
94 and 8 (16) bytes for free list pointers. Thus, the minimum
95 allocatable size is 16/24/32 bytes.
97 Even a request for zero bytes (i.e., malloc(0)) returns a
98 pointer to something of the minimum allocatable size.
100 Maximum allocated size: 4-byte size_t: 2^31 - 8 bytes
101 8-byte size_t: 2^63 - 16 bytes
103 It is assumed that (possibly signed) size_t bit values suffice to
104 represent chunk sizes. `Possibly signed' is due to the fact
105 that `size_t' may be defined on a system as either a signed or
106 an unsigned type. To be conservative, values that would appear
107 as negative numbers are avoided.
108 Requests for sizes with a negative sign bit when the request
109 size is treaded as a long will return null.
111 Maximum overhead wastage per allocated chunk: normally 15 bytes
113 Alignnment demands, plus the minimum allocatable size restriction
114 make the normal worst-case wastage 15 bytes (i.e., up to 15
115 more bytes will be allocated than were requested in malloc), with
117 1. Because requests for zero bytes allocate non-zero space,
118 the worst case wastage for a request of zero bytes is 24 bytes.
119 2. For requests >= mmap_threshold that are serviced via
120 mmap(), the worst case wastage is 8 bytes plus the remainder
121 from a system page (the minimal mmap unit); typically 4096 bytes.
125 Here are some features that are NOT currently supported
127 * No user-definable hooks for callbacks and the like.
128 * No automated mechanism for fully checking that all accesses
129 to malloced memory stay within their bounds.
130 * No support for compaction.
132 * Synopsis of compile-time options:
134 People have reported using previous versions of this malloc on all
135 versions of Unix, sometimes by tweaking some of the defines
136 below. It has been tested most extensively on Solaris and
137 Linux. It is also reported to work on WIN32 platforms.
138 People have also reported adapting this malloc for use in
139 stand-alone embedded systems.
141 The implementation is in straight, hand-tuned ANSI C. Among other
142 consequences, it uses a lot of macros. Because of this, to be at
143 all usable, this code should be compiled using an optimizing compiler
144 (for example gcc -O2) that can simplify expressions and control
147 __STD_C (default: derived from C compiler defines)
148 Nonzero if using ANSI-standard C compiler, a C++ compiler, or
149 a C compiler sufficiently close to ANSI to get away with it.
150 DEBUG (default: NOT defined)
151 Define to enable debugging. Adds fairly extensive assertion-based
152 checking to help track down memory errors, but noticeably slows down
154 REALLOC_ZERO_BYTES_FREES (default: NOT defined)
155 Define this if you think that realloc(p, 0) should be equivalent
156 to free(p). Otherwise, since malloc returns a unique pointer for
157 malloc(0), so does realloc(p, 0).
158 HAVE_MEMCPY (default: defined)
159 Define if you are not otherwise using ANSI STD C, but still
160 have memcpy and memset in your C library and want to use them.
161 Otherwise, simple internal versions are supplied.
162 USE_MEMCPY (default: 1 if HAVE_MEMCPY is defined, 0 otherwise)
163 Define as 1 if you want the C library versions of memset and
164 memcpy called in realloc and calloc (otherwise macro versions are used).
165 At least on some platforms, the simple macro versions usually
166 outperform libc versions.
167 HAVE_MMAP (default: defined as 1)
168 Define to non-zero to optionally make malloc() use mmap() to
169 allocate very large blocks.
170 HAVE_MREMAP (default: defined as 0 unless Linux libc set)
171 Define to non-zero to optionally make realloc() use mremap() to
172 reallocate very large blocks.
173 malloc_getpagesize (default: derived from system #includes)
174 Either a constant or routine call returning the system page size.
175 HAVE_USR_INCLUDE_MALLOC_H (default: NOT defined)
176 Optionally define if you are on a system with a /usr/include/malloc.h
177 that declares struct mallinfo. It is not at all necessary to
178 define this even if you do, but will ensure consistency.
179 INTERNAL_SIZE_T (default: size_t)
180 Define to a 32-bit type (probably `unsigned int') if you are on a
181 64-bit machine, yet do not want or need to allow malloc requests of
182 greater than 2^31 to be handled. This saves space, especially for
184 INTERNAL_LINUX_C_LIB (default: NOT defined)
185 Defined only when compiled as part of Linux libc.
186 Also note that there is some odd internal name-mangling via defines
187 (for example, internally, `malloc' is named `mALLOc') needed
188 when compiling in this case. These look funny but don't otherwise
190 WIN32 (default: undefined)
191 Define this on MS win (95, nt) platforms to compile in sbrk emulation.
192 LACKS_UNISTD_H (default: undefined if not WIN32)
193 Define this if your system does not have a <unistd.h>.
194 LACKS_SYS_PARAM_H (default: undefined if not WIN32)
195 Define this if your system does not have a <sys/param.h>.
196 MORECORE (default: sbrk)
197 The name of the routine to call to obtain more memory from the system.
198 MORECORE_FAILURE (default: -1)
199 The value returned upon failure of MORECORE.
200 MORECORE_CLEARS (default 1)
201 True (1) if the routine mapped to MORECORE zeroes out memory (which
203 DEFAULT_TRIM_THRESHOLD
205 DEFAULT_MMAP_THRESHOLD
207 Default values of tunable parameters (described in detail below)
208 controlling interaction with host system routines (sbrk, mmap, etc).
209 These values may also be changed dynamically via mallopt(). The
210 preset defaults are those that give best performance for typical
212 USE_DL_PREFIX (default: undefined)
213 Prefix all public routines with the string 'dl'. Useful to
214 quickly avoid procedure declaration conflicts and linker symbol
215 conflicts with existing memory allocation routines.
233 #endif /*__cplusplus*/
238 #if (__STD_C || defined(WIN32))
246 #include <stddef.h> /* for size_t */
248 #include <sys/types.h>
255 #include <stdio.h> /* needed for malloc_stats */
266 Because freed chunks may be overwritten with link fields, this
267 malloc will often die when freed memory is overwritten by user
268 programs. This can be very effective (albeit in an annoying way)
269 in helping track down dangling pointers.
271 If you compile with -DDEBUG, a number of assertion checks are
272 enabled that will catch more memory errors. You probably won't be
273 able to make much sense of the actual assertion errors, but they
274 should help you locate incorrectly overwritten memory. The
275 checking is fairly extensive, and will slow down execution
276 noticeably. Calling malloc_stats or mallinfo with DEBUG set will
277 attempt to check every non-mmapped allocated and free chunk in the
278 course of computing the summmaries. (By nature, mmapped regions
279 cannot be checked very much automatically.)
281 Setting DEBUG may also be helpful if you are trying to modify
282 this code. The assertions in the check routines spell out in more
283 detail the assumptions and invariants underlying the algorithms.
290 #define assert(x) ((void)0)
295 INTERNAL_SIZE_T is the word-size used for internal bookkeeping
296 of chunk sizes. On a 64-bit machine, you can reduce malloc
297 overhead by defining INTERNAL_SIZE_T to be a 32 bit `unsigned int'
298 at the expense of not being able to handle requests greater than
299 2^31. This limitation is hardly ever a concern; you are encouraged
300 to set this. However, the default version is the same as size_t.
303 #ifndef INTERNAL_SIZE_T
304 #define INTERNAL_SIZE_T size_t
308 REALLOC_ZERO_BYTES_FREES should be set if a call to
309 realloc with zero bytes should be the same as a call to free.
310 Some people think it should. Otherwise, since this malloc
311 returns a unique pointer for malloc(0), so does realloc(p, 0).
315 /* #define REALLOC_ZERO_BYTES_FREES */
319 WIN32 causes an emulation of sbrk to be compiled in
320 mmap-based options are not currently supported in WIN32.
325 #define MORECORE wsbrk
328 #define LACKS_UNISTD_H
329 #define LACKS_SYS_PARAM_H
332 Include 'windows.h' to get the necessary declarations for the
333 Microsoft Visual C++ data structures and routines used in the 'sbrk'
336 Define WIN32_LEAN_AND_MEAN so that only the essential Microsoft
337 Visual C++ header files are included.
339 #define WIN32_LEAN_AND_MEAN
345 HAVE_MEMCPY should be defined if you are not otherwise using
346 ANSI STD C, but still have memcpy and memset in your C library
347 and want to use them in calloc and realloc. Otherwise simple
348 macro versions are defined here.
350 USE_MEMCPY should be defined as 1 if you actually want to
351 have memset and memcpy called. People report that the macro
352 versions are often enough faster than libc versions on many
353 systems that it is better to use them.
367 #if (__STD_C || defined(HAVE_MEMCPY))
370 void* memset(void*, int, size_t);
371 void* memcpy(void*, const void*, size_t);
374 // On Win32 platforms, 'memset()' and 'memcpy()' are already declared in
385 /* The following macros are only invoked with (2n+1)-multiples of
386 INTERNAL_SIZE_T units, with a positive integer n. This is exploited
387 for fast inline execution when n is small. */
389 #define MALLOC_ZERO(charp, nbytes) \
391 INTERNAL_SIZE_T mzsz = (nbytes); \
392 if(mzsz <= 9*sizeof(mzsz)) { \
393 INTERNAL_SIZE_T* mz = (INTERNAL_SIZE_T*) (charp); \
394 if(mzsz >= 5*sizeof(mzsz)) { *mz++ = 0; \
396 if(mzsz >= 7*sizeof(mzsz)) { *mz++ = 0; \
398 if(mzsz >= 9*sizeof(mzsz)) { *mz++ = 0; \
403 } else memset((charp), 0, mzsz); \
406 #define MALLOC_COPY(dest,src,nbytes) \
408 INTERNAL_SIZE_T mcsz = (nbytes); \
409 if(mcsz <= 9*sizeof(mcsz)) { \
410 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) (src); \
411 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) (dest); \
412 if(mcsz >= 5*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
413 *mcdst++ = *mcsrc++; \
414 if(mcsz >= 7*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
415 *mcdst++ = *mcsrc++; \
416 if(mcsz >= 9*sizeof(mcsz)) { *mcdst++ = *mcsrc++; \
417 *mcdst++ = *mcsrc++; }}} \
418 *mcdst++ = *mcsrc++; \
419 *mcdst++ = *mcsrc++; \
421 } else memcpy(dest, src, mcsz); \
424 #else /* !USE_MEMCPY */
426 /* Use Duff's device for good zeroing/copying performance. */
428 #define MALLOC_ZERO(charp, nbytes) \
430 INTERNAL_SIZE_T* mzp = (INTERNAL_SIZE_T*)(charp); \
431 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
432 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
434 case 0: for(;;) { *mzp++ = 0; \
435 case 7: *mzp++ = 0; \
436 case 6: *mzp++ = 0; \
437 case 5: *mzp++ = 0; \
438 case 4: *mzp++ = 0; \
439 case 3: *mzp++ = 0; \
440 case 2: *mzp++ = 0; \
441 case 1: *mzp++ = 0; if(mcn <= 0) break; mcn--; } \
445 #define MALLOC_COPY(dest,src,nbytes) \
447 INTERNAL_SIZE_T* mcsrc = (INTERNAL_SIZE_T*) src; \
448 INTERNAL_SIZE_T* mcdst = (INTERNAL_SIZE_T*) dest; \
449 long mctmp = (nbytes)/sizeof(INTERNAL_SIZE_T), mcn; \
450 if (mctmp < 8) mcn = 0; else { mcn = (mctmp-1)/8; mctmp %= 8; } \
452 case 0: for(;;) { *mcdst++ = *mcsrc++; \
453 case 7: *mcdst++ = *mcsrc++; \
454 case 6: *mcdst++ = *mcsrc++; \
455 case 5: *mcdst++ = *mcsrc++; \
456 case 4: *mcdst++ = *mcsrc++; \
457 case 3: *mcdst++ = *mcsrc++; \
458 case 2: *mcdst++ = *mcsrc++; \
459 case 1: *mcdst++ = *mcsrc++; if(mcn <= 0) break; mcn--; } \
467 Define HAVE_MMAP to optionally make malloc() use mmap() to
468 allocate very large blocks. These will be returned to the
469 operating system immediately after a free().
477 Define HAVE_MREMAP to make realloc() use mremap() to re-allocate
478 large blocks. This is currently only possible on Linux with
479 kernel versions newer than 1.3.77.
483 #ifdef INTERNAL_LINUX_C_LIB
484 #define HAVE_MREMAP 1
486 #define HAVE_MREMAP 0
494 #include <sys/mman.h>
496 #if !defined(MAP_ANONYMOUS) && defined(MAP_ANON)
497 #define MAP_ANONYMOUS MAP_ANON
500 #endif /* HAVE_MMAP */
503 Access to system page size. To the extent possible, this malloc
504 manages memory from the system in page-size units.
506 The following mechanics for getpagesize were adapted from
507 bsd/gnu getpagesize.h
510 #ifndef LACKS_UNISTD_H
514 #ifndef malloc_getpagesize
515 # ifdef _SC_PAGESIZE /* some SVR4 systems omit an underscore */
516 # ifndef _SC_PAGE_SIZE
517 # define _SC_PAGE_SIZE _SC_PAGESIZE
520 # ifdef _SC_PAGE_SIZE
521 # define malloc_getpagesize sysconf(_SC_PAGE_SIZE)
523 # if defined(BSD) || defined(DGUX) || defined(HAVE_GETPAGESIZE)
524 extern size_t getpagesize();
525 # define malloc_getpagesize getpagesize()
528 # define malloc_getpagesize (4096) /* TBD: Use 'GetSystemInfo' instead */
530 # ifndef LACKS_SYS_PARAM_H
531 # include <sys/param.h>
533 # ifdef EXEC_PAGESIZE
534 # define malloc_getpagesize EXEC_PAGESIZE
538 # define malloc_getpagesize NBPG
540 # define malloc_getpagesize (NBPG * CLSIZE)
544 # define malloc_getpagesize NBPC
547 # define malloc_getpagesize PAGESIZE
549 # define malloc_getpagesize (4096) /* just guess */
563 This version of malloc supports the standard SVID/XPG mallinfo
564 routine that returns a struct containing the same kind of
565 information you can get from malloc_stats. It should work on
566 any SVID/XPG compliant system that has a /usr/include/malloc.h
567 defining struct mallinfo. (If you'd like to install such a thing
568 yourself, cut out the preliminary declarations as described above
569 and below and save them in a malloc.h file. But there's no
570 compelling reason to bother to do this.)
572 The main declaration needed is the mallinfo struct that is returned
573 (by-copy) by mallinfo(). The SVID/XPG malloinfo struct contains a
574 bunch of fields, most of which are not even meaningful in this
575 version of malloc. Some of these fields are are instead filled by
576 mallinfo() with other numbers that might possibly be of interest.
578 HAVE_USR_INCLUDE_MALLOC_H should be set if you have a
579 /usr/include/malloc.h file that includes a declaration of struct
580 mallinfo. If so, it is included; else an SVID2/XPG2 compliant
581 version is declared below. These must be precisely the same for
586 /* #define HAVE_USR_INCLUDE_MALLOC_H */
588 #if HAVE_USR_INCLUDE_MALLOC_H
589 #include "/usr/include/malloc.h"
592 /* SVID2/XPG mallinfo structure */
595 int arena; /* total space allocated from system */
596 int ordblks; /* number of non-inuse chunks */
597 int smblks; /* unused -- always zero */
598 int hblks; /* number of mmapped regions */
599 int hblkhd; /* total space in mmapped regions */
600 int usmblks; /* unused -- always zero */
601 int fsmblks; /* unused -- always zero */
602 int uordblks; /* total allocated space */
603 int fordblks; /* total non-inuse space */
604 int keepcost; /* top-most, releasable (via malloc_trim) space */
607 /* SVID2/XPG mallopt options */
609 #define M_MXFAST 1 /* UNUSED in this malloc */
610 #define M_NLBLKS 2 /* UNUSED in this malloc */
611 #define M_GRAIN 3 /* UNUSED in this malloc */
612 #define M_KEEP 4 /* UNUSED in this malloc */
616 /* mallopt options that actually do something */
618 #define M_TRIM_THRESHOLD -1
620 #define M_MMAP_THRESHOLD -3
621 #define M_MMAP_MAX -4
625 #ifndef DEFAULT_TRIM_THRESHOLD
626 #define DEFAULT_TRIM_THRESHOLD (128 * 1024)
630 M_TRIM_THRESHOLD is the maximum amount of unused top-most memory
631 to keep before releasing via malloc_trim in free().
633 Automatic trimming is mainly useful in long-lived programs.
634 Because trimming via sbrk can be slow on some systems, and can
635 sometimes be wasteful (in cases where programs immediately
636 afterward allocate more large chunks) the value should be high
637 enough so that your overall system performance would improve by
640 The trim threshold and the mmap control parameters (see below)
641 can be traded off with one another. Trimming and mmapping are
642 two different ways of releasing unused memory back to the
643 system. Between these two, it is often possible to keep
644 system-level demands of a long-lived program down to a bare
645 minimum. For example, in one test suite of sessions measuring
646 the XF86 X server on Linux, using a trim threshold of 128K and a
647 mmap threshold of 192K led to near-minimal long term resource
650 If you are using this malloc in a long-lived program, it should
651 pay to experiment with these values. As a rough guide, you
652 might set to a value close to the average size of a process
653 (program) running on your system. Releasing this much memory
654 would allow such a process to run in memory. Generally, it's
655 worth it to tune for trimming rather tham memory mapping when a
656 program undergoes phases where several large chunks are
657 allocated and released in ways that can reuse each other's
658 storage, perhaps mixed with phases where there are no such
659 chunks at all. And in well-behaved long-lived programs,
660 controlling release of large blocks via trimming versus mapping
663 However, in most programs, these parameters serve mainly as
664 protection against the system-level effects of carrying around
665 massive amounts of unneeded memory. Since frequent calls to
666 sbrk, mmap, and munmap otherwise degrade performance, the default
667 parameters are set to relatively high values that serve only as
670 The default trim value is high enough to cause trimming only in
671 fairly extreme (by current memory consumption standards) cases.
672 It must be greater than page size to have any useful effect. To
673 disable trimming completely, you can set to (unsigned long)(-1);
679 #ifndef DEFAULT_TOP_PAD
680 #define DEFAULT_TOP_PAD (0)
684 M_TOP_PAD is the amount of extra `padding' space to allocate or
685 retain whenever sbrk is called. It is used in two ways internally:
687 * When sbrk is called to extend the top of the arena to satisfy
688 a new malloc request, this much padding is added to the sbrk
691 * When malloc_trim is called automatically from free(),
692 it is used as the `pad' argument.
694 In both cases, the actual amount of padding is rounded
695 so that the end of the arena is always a system page boundary.
697 The main reason for using padding is to avoid calling sbrk so
698 often. Having even a small pad greatly reduces the likelihood
699 that nearly every malloc request during program start-up (or
700 after trimming) will invoke sbrk, which needlessly wastes
703 Automatic rounding-up to page-size units is normally sufficient
704 to avoid measurable overhead, so the default is 0. However, in
705 systems where sbrk is relatively slow, it can pay to increase
706 this value, at the expense of carrying around more memory than
712 #ifndef DEFAULT_MMAP_THRESHOLD
713 #define DEFAULT_MMAP_THRESHOLD (128 * 1024)
718 M_MMAP_THRESHOLD is the request size threshold for using mmap()
719 to service a request. Requests of at least this size that cannot
720 be allocated using already-existing space will be serviced via mmap.
721 (If enough normal freed space already exists it is used instead.)
723 Using mmap segregates relatively large chunks of memory so that
724 they can be individually obtained and released from the host
725 system. A request serviced through mmap is never reused by any
726 other request (at least not directly; the system may just so
727 happen to remap successive requests to the same locations).
729 Segregating space in this way has the benefit that mmapped space
730 can ALWAYS be individually released back to the system, which
731 helps keep the system level memory demands of a long-lived
732 program low. Mapped memory can never become `locked' between
733 other chunks, as can happen with normally allocated chunks, which
734 menas that even trimming via malloc_trim would not release them.
736 However, it has the disadvantages that:
738 1. The space cannot be reclaimed, consolidated, and then
739 used to service later requests, as happens with normal chunks.
740 2. It can lead to more wastage because of mmap page alignment
742 3. It causes malloc performance to be more dependent on host
743 system memory management support routines which may vary in
744 implementation quality and may impose arbitrary
745 limitations. Generally, servicing a request via normal
746 malloc steps is faster than going through a system's mmap.
748 All together, these considerations should lead you to use mmap
749 only for relatively large requests.
756 #ifndef DEFAULT_MMAP_MAX
758 #define DEFAULT_MMAP_MAX (64)
760 #define DEFAULT_MMAP_MAX (0)
765 M_MMAP_MAX is the maximum number of requests to simultaneously
766 service using mmap. This parameter exists because:
768 1. Some systems have a limited number of internal tables for
770 2. In most systems, overreliance on mmap can degrade overall
772 3. If a program allocates many large regions, it is probably
773 better off using normal sbrk-based allocation routines that
774 can reclaim and reallocate normal heap memory. Using a
775 small value allows transition into this mode after the
776 first few allocations.
778 Setting to 0 disables all use of mmap. If HAVE_MMAP is not set,
779 the default value is 0, and attempts to set it to non-zero values
780 in mallopt will fail.
787 USE_DL_PREFIX will prefix all public routines with the string 'dl'.
788 Useful to quickly avoid procedure declaration conflicts and linker
789 symbol conflicts with existing memory allocation routines.
793 /* #define USE_DL_PREFIX */
800 Special defines for linux libc
802 Except when compiled using these special defines for Linux libc
803 using weak aliases, this malloc is NOT designed to work in
804 multithreaded applications. No semaphores or other concurrency
805 control are provided to ensure that multiple malloc or free calls
806 don't run at the same time, which could be disasterous. A single
807 semaphore could be used across malloc, realloc, and free (which is
808 essentially the effect of the linux weak alias approach). It would
809 be hard to obtain finer granularity.
814 #ifdef INTERNAL_LINUX_C_LIB
818 Void_t * __default_morecore_init (ptrdiff_t);
819 Void_t *(*__morecore)(ptrdiff_t) = __default_morecore_init;
823 Void_t * __default_morecore_init ();
824 Void_t *(*__morecore)() = __default_morecore_init;
828 #define MORECORE (*__morecore)
829 #define MORECORE_FAILURE 0
830 #define MORECORE_CLEARS 1
832 #else /* INTERNAL_LINUX_C_LIB */
835 extern Void_t* sbrk(ptrdiff_t);
837 extern Void_t* sbrk();
841 #define MORECORE sbrk
844 #ifndef MORECORE_FAILURE
845 #define MORECORE_FAILURE -1
848 #ifndef MORECORE_CLEARS
849 #define MORECORE_CLEARS 1
852 #endif /* INTERNAL_LINUX_C_LIB */
854 #if defined(INTERNAL_LINUX_C_LIB) && defined(__ELF__)
856 #define cALLOc __libc_calloc
857 #define fREe __libc_free
858 #define mALLOc __libc_malloc
859 #define mEMALIGn __libc_memalign
860 #define rEALLOc __libc_realloc
861 #define vALLOc __libc_valloc
862 #define pvALLOc __libc_pvalloc
863 #define mALLINFo __libc_mallinfo
864 #define mALLOPt __libc_mallopt
866 #pragma weak calloc = __libc_calloc
867 #pragma weak free = __libc_free
868 #pragma weak cfree = __libc_free
869 #pragma weak malloc = __libc_malloc
870 #pragma weak memalign = __libc_memalign
871 #pragma weak realloc = __libc_realloc
872 #pragma weak valloc = __libc_valloc
873 #pragma weak pvalloc = __libc_pvalloc
874 #pragma weak mallinfo = __libc_mallinfo
875 #pragma weak mallopt = __libc_mallopt
880 #define cALLOc dlcalloc
882 #define mALLOc dlmalloc
883 #define mEMALIGn dlmemalign
884 #define rEALLOc dlrealloc
885 #define vALLOc dlvalloc
886 #define pvALLOc dlpvalloc
887 #define mALLINFo dlmallinfo
888 #define mALLOPt dlmallopt
889 #else /* USE_DL_PREFIX */
890 #define cALLOc calloc
892 #define mALLOc malloc
893 #define mEMALIGn memalign
894 #define rEALLOc realloc
895 #define vALLOc valloc
896 #define pvALLOc pvalloc
897 #define mALLINFo mallinfo
898 #define mALLOPt mallopt
899 #endif /* USE_DL_PREFIX */
903 /* Public routines */
907 Void_t* mALLOc(size_t);
909 Void_t* rEALLOc(Void_t*, size_t);
910 Void_t* mEMALIGn(size_t, size_t);
911 Void_t* vALLOc(size_t);
912 Void_t* pvALLOc(size_t);
913 Void_t* cALLOc(size_t, size_t);
915 int malloc_trim(size_t);
916 size_t malloc_usable_size(Void_t*);
918 int mALLOPt(int, int);
919 struct mallinfo mALLINFo(void);
930 size_t malloc_usable_size();
933 struct mallinfo mALLINFo();
938 }; /* end of extern "C" */
941 /* ---------- To make a malloc.h, end cutting here ------------ */
945 Emulation of sbrk for WIN32
946 All code within the ifdef WIN32 is untested by me.
948 Thanks to Martin Fong and others for supplying this.
954 #define AlignPage(add) (((add) + (malloc_getpagesize-1)) & \
955 ~(malloc_getpagesize-1))
956 #define AlignPage64K(add) (((add) + (0x10000 - 1)) & ~(0x10000 - 1))
958 /* resrve 64MB to insure large contiguous space */
959 #define RESERVED_SIZE (1024*1024*64)
960 #define NEXT_SIZE (2048*1024)
961 #define TOP_MEMORY ((unsigned long)2*1024*1024*1024)
963 struct GmListElement;
964 typedef struct GmListElement GmListElement;
972 static GmListElement* head = 0;
973 static unsigned int gNextAddress = 0;
974 static unsigned int gAddressBase = 0;
975 static unsigned int gAllocatedSize = 0;
978 GmListElement* makeGmListElement (void* bas)
981 this = (GmListElement*)(void*)LocalAlloc (0, sizeof (GmListElement));
995 assert ( (head == NULL) || (head->base == (void*)gAddressBase));
996 if (gAddressBase && (gNextAddress - gAddressBase))
998 rval = VirtualFree ((void*)gAddressBase,
999 gNextAddress - gAddressBase,
1005 GmListElement* next = head->next;
1006 rval = VirtualFree (head->base, 0, MEM_RELEASE);
1014 void* findRegion (void* start_address, unsigned long size)
1016 MEMORY_BASIC_INFORMATION info;
1017 if (size >= TOP_MEMORY) return NULL;
1019 while ((unsigned long)start_address + size < TOP_MEMORY)
1021 VirtualQuery (start_address, &info, sizeof (info));
1022 if ((info.State == MEM_FREE) && (info.RegionSize >= size))
1023 return start_address;
1026 // Requested region is not available so see if the
1027 // next region is available. Set 'start_address'
1028 // to the next region and call 'VirtualQuery()'
1031 start_address = (char*)info.BaseAddress + info.RegionSize;
1033 // Make sure we start looking for the next region
1034 // on the *next* 64K boundary. Otherwise, even if
1035 // the new region is free according to
1036 // 'VirtualQuery()', the subsequent call to
1037 // 'VirtualAlloc()' (which follows the call to
1038 // this routine in 'wsbrk()') will round *down*
1039 // the requested address to a 64K boundary which
1040 // we already know is an address in the
1041 // unavailable region. Thus, the subsequent call
1042 // to 'VirtualAlloc()' will fail and bring us back
1043 // here, causing us to go into an infinite loop.
1046 (void *) AlignPage64K((unsigned long) start_address);
1054 void* wsbrk (long size)
1059 if (gAddressBase == 0)
1061 gAllocatedSize = max (RESERVED_SIZE, AlignPage (size));
1062 gNextAddress = gAddressBase =
1063 (unsigned int)VirtualAlloc (NULL, gAllocatedSize,
1064 MEM_RESERVE, PAGE_NOACCESS);
1065 } else if (AlignPage (gNextAddress + size) > (gAddressBase +
1068 long new_size = max (NEXT_SIZE, AlignPage (size));
1069 void* new_address = (void*)(gAddressBase+gAllocatedSize);
1072 new_address = findRegion (new_address, new_size);
1074 if (new_address == 0)
1077 gAddressBase = gNextAddress =
1078 (unsigned int)VirtualAlloc (new_address, new_size,
1079 MEM_RESERVE, PAGE_NOACCESS);
1080 // repeat in case of race condition
1081 // The region that we found has been snagged
1082 // by another thread
1084 while (gAddressBase == 0);
1086 assert (new_address == (void*)gAddressBase);
1088 gAllocatedSize = new_size;
1090 if (!makeGmListElement ((void*)gAddressBase))
1093 if ((size + gNextAddress) > AlignPage (gNextAddress))
1096 res = VirtualAlloc ((void*)AlignPage (gNextAddress),
1097 (size + gNextAddress -
1098 AlignPage (gNextAddress)),
1099 MEM_COMMIT, PAGE_READWRITE);
1103 tmp = (void*)gNextAddress;
1104 gNextAddress = (unsigned int)tmp + size;
1109 unsigned int alignedGoal = AlignPage (gNextAddress + size);
1110 /* Trim by releasing the virtual memory */
1111 if (alignedGoal >= gAddressBase)
1113 VirtualFree ((void*)alignedGoal, gNextAddress - alignedGoal,
1115 gNextAddress = gNextAddress + size;
1116 return (void*)gNextAddress;
1120 VirtualFree ((void*)gAddressBase, gNextAddress - gAddressBase,
1122 gNextAddress = gAddressBase;
1128 return (void*)gNextAddress;
1143 INTERNAL_SIZE_T prev_size; /* Size of previous chunk (if free). */
1144 INTERNAL_SIZE_T size; /* Size in bytes, including overhead. */
1145 struct malloc_chunk* fd; /* double links -- used only if free. */
1146 struct malloc_chunk* bk;
1149 typedef struct malloc_chunk* mchunkptr;
1153 malloc_chunk details:
1155 (The following includes lightly edited explanations by Colin Plumb.)
1157 Chunks of memory are maintained using a `boundary tag' method as
1158 described in e.g., Knuth or Standish. (See the paper by Paul
1159 Wilson ftp://ftp.cs.utexas.edu/pub/garbage/allocsrv.ps for a
1160 survey of such techniques.) Sizes of free chunks are stored both
1161 in the front of each chunk and at the end. This makes
1162 consolidating fragmented chunks into bigger chunks very fast. The
1163 size fields also hold bits representing whether chunks are free or
1166 An allocated chunk looks like this:
1169 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1170 | Size of previous chunk, if allocated | |
1171 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1172 | Size of chunk, in bytes |P|
1173 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1174 | User data starts here... .
1176 . (malloc_usable_space() bytes) .
1178 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1180 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1183 Where "chunk" is the front of the chunk for the purpose of most of
1184 the malloc code, but "mem" is the pointer that is returned to the
1185 user. "Nextchunk" is the beginning of the next contiguous chunk.
1187 Chunks always begin on even word boundries, so the mem portion
1188 (which is returned to the user) is also on an even word boundary, and
1189 thus double-word aligned.
1191 Free chunks are stored in circular doubly-linked lists, and look like this:
1193 chunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1194 | Size of previous chunk |
1195 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1196 `head:' | Size of chunk, in bytes |P|
1197 mem-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1198 | Forward pointer to next chunk in list |
1199 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1200 | Back pointer to previous chunk in list |
1201 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1202 | Unused space (may be 0 bytes long) .
1205 nextchunk-> +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1206 `foot:' | Size of chunk, in bytes |
1207 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
1209 The P (PREV_INUSE) bit, stored in the unused low-order bit of the
1210 chunk size (which is always a multiple of two words), is an in-use
1211 bit for the *previous* chunk. If that bit is *clear*, then the
1212 word before the current chunk size contains the previous chunk
1213 size, and can be used to find the front of the previous chunk.
1214 (The very first chunk allocated always has this bit set,
1215 preventing access to non-existent (or non-owned) memory.)
1217 Note that the `foot' of the current chunk is actually represented
1218 as the prev_size of the NEXT chunk. (This makes it easier to
1219 deal with alignments etc).
1221 The two exceptions to all this are
1223 1. The special chunk `top', which doesn't bother using the
1224 trailing size field since there is no
1225 next contiguous chunk that would have to index off it. (After
1226 initialization, `top' is forced to always exist. If it would
1227 become less than MINSIZE bytes long, it is replenished via
1230 2. Chunks allocated via mmap, which have the second-lowest-order
1231 bit (IS_MMAPPED) set in their size fields. Because they are
1232 never merged or traversed from any other chunk, they have no
1233 foot size or inuse information.
1235 Available chunks are kept in any of several places (all declared below):
1237 * `av': An array of chunks serving as bin headers for consolidated
1238 chunks. Each bin is doubly linked. The bins are approximately
1239 proportionally (log) spaced. There are a lot of these bins
1240 (128). This may look excessive, but works very well in
1241 practice. All procedures maintain the invariant that no
1242 consolidated chunk physically borders another one. Chunks in
1243 bins are kept in size order, with ties going to the
1244 approximately least recently used chunk.
1246 The chunks in each bin are maintained in decreasing sorted order by
1247 size. This is irrelevant for the small bins, which all contain
1248 the same-sized chunks, but facilitates best-fit allocation for
1249 larger chunks. (These lists are just sequential. Keeping them in
1250 order almost never requires enough traversal to warrant using
1251 fancier ordered data structures.) Chunks of the same size are
1252 linked with the most recently freed at the front, and allocations
1253 are taken from the back. This results in LRU or FIFO allocation
1254 order, which tends to give each chunk an equal opportunity to be
1255 consolidated with adjacent freed chunks, resulting in larger free
1256 chunks and less fragmentation.
1258 * `top': The top-most available chunk (i.e., the one bordering the
1259 end of available memory) is treated specially. It is never
1260 included in any bin, is used only if no other chunk is
1261 available, and is released back to the system if it is very
1262 large (see M_TRIM_THRESHOLD).
1264 * `last_remainder': A bin holding only the remainder of the
1265 most recently split (non-top) chunk. This bin is checked
1266 before other non-fitting chunks, so as to provide better
1267 locality for runs of sequentially allocated chunks.
1269 * Implicitly, through the host system's memory mapping tables.
1270 If supported, requests greater than a threshold are usually
1271 serviced via calls to mmap, and then later released via munmap.
1280 /* sizes, alignments */
1282 #define SIZE_SZ (sizeof(INTERNAL_SIZE_T))
1283 #define MALLOC_ALIGNMENT (SIZE_SZ + SIZE_SZ)
1284 #define MALLOC_ALIGN_MASK (MALLOC_ALIGNMENT - 1)
1285 #define MINSIZE (sizeof(struct malloc_chunk))
1287 /* conversion from malloc headers to user pointers, and back */
1289 #define chunk2mem(p) ((Void_t*)((char*)(p) + 2*SIZE_SZ))
1290 #define mem2chunk(mem) ((mchunkptr)((char*)(mem) - 2*SIZE_SZ))
1292 /* pad request bytes into a usable size */
1294 #define request2size(req) \
1295 (((long)((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) < \
1296 (long)(MINSIZE + MALLOC_ALIGN_MASK)) ? MINSIZE : \
1297 (((req) + (SIZE_SZ + MALLOC_ALIGN_MASK)) & ~(MALLOC_ALIGN_MASK)))
1299 /* Check if m has acceptable alignment */
1301 #define aligned_OK(m) (((unsigned long)((m)) & (MALLOC_ALIGN_MASK)) == 0)
1307 Physical chunk operations
1311 /* size field is or'ed with PREV_INUSE when previous adjacent chunk in use */
1313 #define PREV_INUSE 0x1
1315 /* size field is or'ed with IS_MMAPPED if the chunk was obtained with mmap() */
1317 #define IS_MMAPPED 0x2
1319 /* Bits to mask off when extracting size */
1321 #define SIZE_BITS (PREV_INUSE|IS_MMAPPED)
1324 /* Ptr to next physical malloc_chunk. */
1326 #define next_chunk(p) ((mchunkptr)( ((char*)(p)) + ((p)->size & ~PREV_INUSE) ))
1328 /* Ptr to previous physical malloc_chunk */
1330 #define prev_chunk(p)\
1331 ((mchunkptr)( ((char*)(p)) - ((p)->prev_size) ))
1334 /* Treat space at ptr + offset as a chunk */
1336 #define chunk_at_offset(p, s) ((mchunkptr)(((char*)(p)) + (s)))
1342 Dealing with use bits
1345 /* extract p's inuse bit */
1348 ((((mchunkptr)(((char*)(p))+((p)->size & ~PREV_INUSE)))->size) & PREV_INUSE)
1350 /* extract inuse bit of previous chunk */
1352 #define prev_inuse(p) ((p)->size & PREV_INUSE)
1354 /* check for mmap()'ed chunk */
1356 #define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED)
1358 /* set/clear chunk as in use without otherwise disturbing */
1360 #define set_inuse(p)\
1361 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size |= PREV_INUSE
1363 #define clear_inuse(p)\
1364 ((mchunkptr)(((char*)(p)) + ((p)->size & ~PREV_INUSE)))->size &= ~(PREV_INUSE)
1366 /* check/set/clear inuse bits in known places */
1368 #define inuse_bit_at_offset(p, s)\
1369 (((mchunkptr)(((char*)(p)) + (s)))->size & PREV_INUSE)
1371 #define set_inuse_bit_at_offset(p, s)\
1372 (((mchunkptr)(((char*)(p)) + (s)))->size |= PREV_INUSE)
1374 #define clear_inuse_bit_at_offset(p, s)\
1375 (((mchunkptr)(((char*)(p)) + (s)))->size &= ~(PREV_INUSE))
1381 Dealing with size fields
1384 /* Get size, ignoring use bits */
1386 #define chunksize(p) ((p)->size & ~(SIZE_BITS))
1388 /* Set size at head, without disturbing its use bit */
1390 #define set_head_size(p, s) ((p)->size = (((p)->size & PREV_INUSE) | (s)))
1392 /* Set size/use ignoring previous bits in header */
1394 #define set_head(p, s) ((p)->size = (s))
1396 /* Set size at footer (only when chunk is not in use) */
1398 #define set_foot(p, s) (((mchunkptr)((char*)(p) + (s)))->prev_size = (s))
1407 The bins, `av_' are an array of pairs of pointers serving as the
1408 heads of (initially empty) doubly-linked lists of chunks, laid out
1409 in a way so that each pair can be treated as if it were in a
1410 malloc_chunk. (This way, the fd/bk offsets for linking bin heads
1411 and chunks are the same).
1413 Bins for sizes < 512 bytes contain chunks of all the same size, spaced
1414 8 bytes apart. Larger bins are approximately logarithmically
1415 spaced. (See the table below.) The `av_' array is never mentioned
1416 directly in the code, but instead via bin access macros.
1424 4 bins of size 32768
1425 2 bins of size 262144
1426 1 bin of size what's left
1428 There is actually a little bit of slop in the numbers in bin_index
1429 for the sake of speed. This makes no difference elsewhere.
1431 The special chunks `top' and `last_remainder' get their own bins,
1432 (this is implemented via yet more trickery with the av_ array),
1433 although `top' is never properly linked to its bin since it is
1434 always handled specially.
1438 #define NAV 128 /* number of bins */
1440 typedef struct malloc_chunk* mbinptr;
1444 #define bin_at(i) ((mbinptr)((char*)&(av_[2*(i) + 2]) - 2*SIZE_SZ))
1445 #define next_bin(b) ((mbinptr)((char*)(b) + 2 * sizeof(mbinptr)))
1446 #define prev_bin(b) ((mbinptr)((char*)(b) - 2 * sizeof(mbinptr)))
1449 The first 2 bins are never indexed. The corresponding av_ cells are instead
1450 used for bookkeeping. This is not to save space, but to simplify
1451 indexing, maintain locality, and avoid some initialization tests.
1454 #define top (bin_at(0)->fd) /* The topmost chunk */
1455 #define last_remainder (bin_at(1)) /* remainder from last split */
1459 Because top initially points to its own bin with initial
1460 zero size, thus forcing extension on the first malloc request,
1461 we avoid having any special code in malloc to check whether
1462 it even exists yet. But we still need to in malloc_extend_top.
1465 #define initial_top ((mchunkptr)(bin_at(0)))
1467 /* Helper macro to initialize bins */
1469 #define IAV(i) bin_at(i), bin_at(i)
1471 static mbinptr av_[NAV * 2 + 2] = {
1473 IAV(0), IAV(1), IAV(2), IAV(3), IAV(4), IAV(5), IAV(6), IAV(7),
1474 IAV(8), IAV(9), IAV(10), IAV(11), IAV(12), IAV(13), IAV(14), IAV(15),
1475 IAV(16), IAV(17), IAV(18), IAV(19), IAV(20), IAV(21), IAV(22), IAV(23),
1476 IAV(24), IAV(25), IAV(26), IAV(27), IAV(28), IAV(29), IAV(30), IAV(31),
1477 IAV(32), IAV(33), IAV(34), IAV(35), IAV(36), IAV(37), IAV(38), IAV(39),
1478 IAV(40), IAV(41), IAV(42), IAV(43), IAV(44), IAV(45), IAV(46), IAV(47),
1479 IAV(48), IAV(49), IAV(50), IAV(51), IAV(52), IAV(53), IAV(54), IAV(55),
1480 IAV(56), IAV(57), IAV(58), IAV(59), IAV(60), IAV(61), IAV(62), IAV(63),
1481 IAV(64), IAV(65), IAV(66), IAV(67), IAV(68), IAV(69), IAV(70), IAV(71),
1482 IAV(72), IAV(73), IAV(74), IAV(75), IAV(76), IAV(77), IAV(78), IAV(79),
1483 IAV(80), IAV(81), IAV(82), IAV(83), IAV(84), IAV(85), IAV(86), IAV(87),
1484 IAV(88), IAV(89), IAV(90), IAV(91), IAV(92), IAV(93), IAV(94), IAV(95),
1485 IAV(96), IAV(97), IAV(98), IAV(99), IAV(100), IAV(101), IAV(102), IAV(103),
1486 IAV(104), IAV(105), IAV(106), IAV(107), IAV(108), IAV(109), IAV(110), IAV(111),
1487 IAV(112), IAV(113), IAV(114), IAV(115), IAV(116), IAV(117), IAV(118), IAV(119),
1488 IAV(120), IAV(121), IAV(122), IAV(123), IAV(124), IAV(125), IAV(126), IAV(127)
1493 /* field-extraction macros */
1495 #define first(b) ((b)->fd)
1496 #define last(b) ((b)->bk)
1502 #define bin_index(sz) \
1503 (((((unsigned long)(sz)) >> 9) == 0) ? (((unsigned long)(sz)) >> 3): \
1504 ((((unsigned long)(sz)) >> 9) <= 4) ? 56 + (((unsigned long)(sz)) >> 6): \
1505 ((((unsigned long)(sz)) >> 9) <= 20) ? 91 + (((unsigned long)(sz)) >> 9): \
1506 ((((unsigned long)(sz)) >> 9) <= 84) ? 110 + (((unsigned long)(sz)) >> 12): \
1507 ((((unsigned long)(sz)) >> 9) <= 340) ? 119 + (((unsigned long)(sz)) >> 15): \
1508 ((((unsigned long)(sz)) >> 9) <= 1364) ? 124 + (((unsigned long)(sz)) >> 18): \
1511 bins for chunks < 512 are all spaced 8 bytes apart, and hold
1512 identically sized chunks. This is exploited in malloc.
1515 #define MAX_SMALLBIN 63
1516 #define MAX_SMALLBIN_SIZE 512
1517 #define SMALLBIN_WIDTH 8
1519 #define smallbin_index(sz) (((unsigned long)(sz)) >> 3)
1522 Requests are `small' if both the corresponding and the next bin are small
1525 #define is_small_request(nb) (nb < MAX_SMALLBIN_SIZE - SMALLBIN_WIDTH)
1530 To help compensate for the large number of bins, a one-level index
1531 structure is used for bin-by-bin searching. `binblocks' is a
1532 one-word bitvector recording whether groups of BINBLOCKWIDTH bins
1533 have any (possibly) non-empty bins, so they can be skipped over
1534 all at once during during traversals. The bits are NOT always
1535 cleared as soon as all bins in a block are empty, but instead only
1536 when all are noticed to be empty during traversal in malloc.
1539 #define BINBLOCKWIDTH 4 /* bins per block */
1541 #define binblocks (bin_at(0)->size) /* bitvector of nonempty blocks */
1543 /* bin<->block macros */
1545 #define idx2binblock(ix) ((unsigned)1 << (ix / BINBLOCKWIDTH))
1546 #define mark_binblock(ii) (binblocks |= idx2binblock(ii))
1547 #define clear_binblock(ii) (binblocks &= ~(idx2binblock(ii)))
1553 /* Other static bookkeeping data */
1555 /* variables holding tunable values */
1557 static unsigned long trim_threshold = DEFAULT_TRIM_THRESHOLD;
1558 static unsigned long top_pad = DEFAULT_TOP_PAD;
1559 static unsigned int n_mmaps_max = DEFAULT_MMAP_MAX;
1560 static unsigned long mmap_threshold = DEFAULT_MMAP_THRESHOLD;
1562 /* The first value returned from sbrk */
1563 static char* sbrk_base = (char*)(-1);
1565 /* The maximum memory obtained from system via sbrk */
1566 static unsigned long max_sbrked_mem = 0;
1568 /* The maximum via either sbrk or mmap */
1569 static unsigned long max_total_mem = 0;
1571 /* internal working copy of mallinfo */
1572 static struct mallinfo current_mallinfo = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };
1574 /* The total memory obtained from system via sbrk */
1575 #define sbrked_mem (current_mallinfo.arena)
1577 /* Tracking mmaps */
1579 static unsigned int n_mmaps = 0;
1580 static unsigned int max_n_mmaps = 0;
1581 static unsigned long mmapped_mem = 0;
1582 static unsigned long max_mmapped_mem = 0;
1594 These routines make a number of assertions about the states
1595 of data structures that should be true at all times. If any
1596 are not true, it's very likely that a user program has somehow
1597 trashed memory. (It's also possible that there is a coding error
1598 in malloc. In which case, please report it!)
1602 static void do_check_chunk(mchunkptr p)
1604 static void do_check_chunk(p) mchunkptr p;
1607 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1609 /* No checkable chunk is mmapped */
1610 assert(!chunk_is_mmapped(p));
1612 /* Check for legal address ... */
1613 assert((char*)p >= sbrk_base);
1615 assert((char*)p + sz <= (char*)top);
1617 assert((char*)p + sz <= sbrk_base + sbrked_mem);
1623 static void do_check_free_chunk(mchunkptr p)
1625 static void do_check_free_chunk(p) mchunkptr p;
1628 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1629 mchunkptr next = chunk_at_offset(p, sz);
1633 /* Check whether it claims to be free ... */
1636 /* Unless a special marker, must have OK fields */
1637 if ((long)sz >= (long)MINSIZE)
1639 assert((sz & MALLOC_ALIGN_MASK) == 0);
1640 assert(aligned_OK(chunk2mem(p)));
1641 /* ... matching footer field */
1642 assert(next->prev_size == sz);
1643 /* ... and is fully consolidated */
1644 assert(prev_inuse(p));
1645 assert (next == top || inuse(next));
1647 /* ... and has minimally sane links */
1648 assert(p->fd->bk == p);
1649 assert(p->bk->fd == p);
1651 else /* markers are always of size SIZE_SZ */
1652 assert(sz == SIZE_SZ);
1656 static void do_check_inuse_chunk(mchunkptr p)
1658 static void do_check_inuse_chunk(p) mchunkptr p;
1661 mchunkptr next = next_chunk(p);
1664 /* Check whether it claims to be in use ... */
1667 /* ... and is surrounded by OK chunks.
1668 Since more things can be checked with free chunks than inuse ones,
1669 if an inuse chunk borders them and debug is on, it's worth doing them.
1673 mchunkptr prv = prev_chunk(p);
1674 assert(next_chunk(prv) == p);
1675 do_check_free_chunk(prv);
1679 assert(prev_inuse(next));
1680 assert(chunksize(next) >= MINSIZE);
1682 else if (!inuse(next))
1683 do_check_free_chunk(next);
1688 static void do_check_malloced_chunk(mchunkptr p, INTERNAL_SIZE_T s)
1690 static void do_check_malloced_chunk(p, s) mchunkptr p; INTERNAL_SIZE_T s;
1693 INTERNAL_SIZE_T sz = p->size & ~PREV_INUSE;
1696 do_check_inuse_chunk(p);
1698 /* Legal size ... */
1699 assert((long)sz >= (long)MINSIZE);
1700 assert((sz & MALLOC_ALIGN_MASK) == 0);
1702 assert(room < (long)MINSIZE);
1704 /* ... and alignment */
1705 assert(aligned_OK(chunk2mem(p)));
1708 /* ... and was allocated at front of an available chunk */
1709 assert(prev_inuse(p));
1714 #define check_free_chunk(P) do_check_free_chunk(P)
1715 #define check_inuse_chunk(P) do_check_inuse_chunk(P)
1716 #define check_chunk(P) do_check_chunk(P)
1717 #define check_malloced_chunk(P,N) do_check_malloced_chunk(P,N)
1719 #define check_free_chunk(P)
1720 #define check_inuse_chunk(P)
1721 #define check_chunk(P)
1722 #define check_malloced_chunk(P,N)
1728 Macro-based internal utilities
1733 Linking chunks in bin lists.
1734 Call these only with variables, not arbitrary expressions, as arguments.
1738 Place chunk p of size s in its bin, in size order,
1739 putting it ahead of others of same size.
1743 #define frontlink(P, S, IDX, BK, FD) \
1745 if (S < MAX_SMALLBIN_SIZE) \
1747 IDX = smallbin_index(S); \
1748 mark_binblock(IDX); \
1753 FD->bk = BK->fd = P; \
1757 IDX = bin_index(S); \
1760 if (FD == BK) mark_binblock(IDX); \
1763 while (FD != BK && S < chunksize(FD)) FD = FD->fd; \
1768 FD->bk = BK->fd = P; \
1773 /* take a chunk off a list */
1775 #define unlink(P, BK, FD) \
1783 /* Place p as the last remainder */
1785 #define link_last_remainder(P) \
1787 last_remainder->fd = last_remainder->bk = P; \
1788 P->fd = P->bk = last_remainder; \
1791 /* Clear the last_remainder bin */
1793 #define clear_last_remainder \
1794 (last_remainder->fd = last_remainder->bk = last_remainder)
1801 /* Routines dealing with mmap(). */
1806 static mchunkptr mmap_chunk(size_t size)
1808 static mchunkptr mmap_chunk(size) size_t size;
1811 size_t page_mask = malloc_getpagesize - 1;
1814 #ifndef MAP_ANONYMOUS
1818 if(n_mmaps >= n_mmaps_max) return 0; /* too many regions */
1820 /* For mmapped chunks, the overhead is one SIZE_SZ unit larger, because
1821 * there is no following chunk whose prev_size field could be used.
1823 size = (size + SIZE_SZ + page_mask) & ~page_mask;
1825 #ifdef MAP_ANONYMOUS
1826 p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE,
1827 MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
1828 #else /* !MAP_ANONYMOUS */
1831 fd = open("/dev/zero", O_RDWR);
1832 if(fd < 0) return 0;
1834 p = (mchunkptr)mmap(0, size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
1837 if(p == (mchunkptr)-1) return 0;
1840 if (n_mmaps > max_n_mmaps) max_n_mmaps = n_mmaps;
1842 /* We demand that eight bytes into a page must be 8-byte aligned. */
1843 assert(aligned_OK(chunk2mem(p)));
1845 /* The offset to the start of the mmapped region is stored
1846 * in the prev_size field of the chunk; normally it is zero,
1847 * but that can be changed in memalign().
1850 set_head(p, size|IS_MMAPPED);
1852 mmapped_mem += size;
1853 if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1854 max_mmapped_mem = mmapped_mem;
1855 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1856 max_total_mem = mmapped_mem + sbrked_mem;
1861 static void munmap_chunk(mchunkptr p)
1863 static void munmap_chunk(p) mchunkptr p;
1866 INTERNAL_SIZE_T size = chunksize(p);
1869 assert (chunk_is_mmapped(p));
1870 assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1871 assert((n_mmaps > 0));
1872 assert(((p->prev_size + size) & (malloc_getpagesize-1)) == 0);
1875 mmapped_mem -= (size + p->prev_size);
1877 ret = munmap((char *)p - p->prev_size, size + p->prev_size);
1879 /* munmap returns non-zero on failure */
1886 static mchunkptr mremap_chunk(mchunkptr p, size_t new_size)
1888 static mchunkptr mremap_chunk(p, new_size) mchunkptr p; size_t new_size;
1891 size_t page_mask = malloc_getpagesize - 1;
1892 INTERNAL_SIZE_T offset = p->prev_size;
1893 INTERNAL_SIZE_T size = chunksize(p);
1896 assert (chunk_is_mmapped(p));
1897 assert(! ((char*)p >= sbrk_base && (char*)p < sbrk_base + sbrked_mem));
1898 assert((n_mmaps > 0));
1899 assert(((size + offset) & (malloc_getpagesize-1)) == 0);
1901 /* Note the extra SIZE_SZ overhead as in mmap_chunk(). */
1902 new_size = (new_size + offset + SIZE_SZ + page_mask) & ~page_mask;
1904 cp = (char *)mremap((char *)p - offset, size + offset, new_size, 1);
1906 if (cp == (char *)-1) return 0;
1908 p = (mchunkptr)(cp + offset);
1910 assert(aligned_OK(chunk2mem(p)));
1912 assert((p->prev_size == offset));
1913 set_head(p, (new_size - offset)|IS_MMAPPED);
1915 mmapped_mem -= size + offset;
1916 mmapped_mem += new_size;
1917 if ((unsigned long)mmapped_mem > (unsigned long)max_mmapped_mem)
1918 max_mmapped_mem = mmapped_mem;
1919 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
1920 max_total_mem = mmapped_mem + sbrked_mem;
1924 #endif /* HAVE_MREMAP */
1926 #endif /* HAVE_MMAP */
1932 Extend the top-most chunk by obtaining memory from system.
1933 Main interface to sbrk (but see also malloc_trim).
1937 static void malloc_extend_top(INTERNAL_SIZE_T nb)
1939 static void malloc_extend_top(nb) INTERNAL_SIZE_T nb;
1942 char* brk; /* return value from sbrk */
1943 INTERNAL_SIZE_T front_misalign; /* unusable bytes at front of sbrked space */
1944 INTERNAL_SIZE_T correction; /* bytes for 2nd sbrk call */
1945 char* new_brk; /* return of 2nd sbrk call */
1946 INTERNAL_SIZE_T top_size; /* new size of top chunk */
1948 mchunkptr old_top = top; /* Record state of old top */
1949 INTERNAL_SIZE_T old_top_size = chunksize(old_top);
1950 char* old_end = (char*)(chunk_at_offset(old_top, old_top_size));
1952 /* Pad request with top_pad plus minimal overhead */
1954 INTERNAL_SIZE_T sbrk_size = nb + top_pad + MINSIZE;
1955 unsigned long pagesz = malloc_getpagesize;
1957 /* If not the first time through, round to preserve page boundary */
1958 /* Otherwise, we need to correct to a page size below anyway. */
1959 /* (We also correct below if an intervening foreign sbrk call.) */
1961 if (sbrk_base != (char*)(-1))
1962 sbrk_size = (sbrk_size + (pagesz - 1)) & ~(pagesz - 1);
1964 brk = (char*)(MORECORE (sbrk_size));
1966 /* Fail if sbrk failed or if a foreign sbrk call killed our space */
1967 if (brk == (char*)(MORECORE_FAILURE) ||
1968 (brk < old_end && old_top != initial_top))
1971 sbrked_mem += sbrk_size;
1973 if (brk == old_end) /* can just add bytes to current top */
1975 top_size = sbrk_size + old_top_size;
1976 set_head(top, top_size | PREV_INUSE);
1980 if (sbrk_base == (char*)(-1)) /* First time through. Record base */
1982 else /* Someone else called sbrk(). Count those bytes as sbrked_mem. */
1983 sbrked_mem += brk - (char*)old_end;
1985 /* Guarantee alignment of first new chunk made from this space */
1986 front_misalign = (unsigned long)chunk2mem(brk) & MALLOC_ALIGN_MASK;
1987 if (front_misalign > 0)
1989 correction = (MALLOC_ALIGNMENT) - front_misalign;
1995 /* Guarantee the next brk will be at a page boundary */
1997 correction += ((((unsigned long)(brk + sbrk_size))+(pagesz-1)) &
1998 ~(pagesz - 1)) - ((unsigned long)(brk + sbrk_size));
2000 /* Allocate correction */
2001 new_brk = (char*)(MORECORE (correction));
2002 if (new_brk == (char*)(MORECORE_FAILURE)) return;
2004 sbrked_mem += correction;
2006 top = (mchunkptr)brk;
2007 top_size = new_brk - brk + correction;
2008 set_head(top, top_size | PREV_INUSE);
2010 if (old_top != initial_top)
2013 /* There must have been an intervening foreign sbrk call. */
2014 /* A double fencepost is necessary to prevent consolidation */
2016 /* If not enough space to do this, then user did something very wrong */
2017 if (old_top_size < MINSIZE)
2019 set_head(top, PREV_INUSE); /* will force null return from malloc */
2023 /* Also keep size a multiple of MALLOC_ALIGNMENT */
2024 old_top_size = (old_top_size - 3*SIZE_SZ) & ~MALLOC_ALIGN_MASK;
2025 set_head_size(old_top, old_top_size);
2026 chunk_at_offset(old_top, old_top_size )->size =
2028 chunk_at_offset(old_top, old_top_size + SIZE_SZ)->size =
2030 /* If possible, release the rest. */
2031 if (old_top_size >= MINSIZE)
2032 fREe(chunk2mem(old_top));
2036 if ((unsigned long)sbrked_mem > (unsigned long)max_sbrked_mem)
2037 max_sbrked_mem = sbrked_mem;
2038 if ((unsigned long)(mmapped_mem + sbrked_mem) > (unsigned long)max_total_mem)
2039 max_total_mem = mmapped_mem + sbrked_mem;
2041 /* We always land on a page boundary */
2042 assert(((unsigned long)((char*)top + top_size) & (pagesz - 1)) == 0);
2048 /* Main public routines */
2054 The requested size is first converted into a usable form, `nb'.
2055 This currently means to add 4 bytes overhead plus possibly more to
2056 obtain 8-byte alignment and/or to obtain a size of at least
2057 MINSIZE (currently 16 bytes), the smallest allocatable size.
2058 (All fits are considered `exact' if they are within MINSIZE bytes.)
2060 From there, the first successful of the following steps is taken:
2062 1. The bin corresponding to the request size is scanned, and if
2063 a chunk of exactly the right size is found, it is taken.
2065 2. The most recently remaindered chunk is used if it is big
2066 enough. This is a form of (roving) first fit, used only in
2067 the absence of exact fits. Runs of consecutive requests use
2068 the remainder of the chunk used for the previous such request
2069 whenever possible. This limited use of a first-fit style
2070 allocation strategy tends to give contiguous chunks
2071 coextensive lifetimes, which improves locality and can reduce
2072 fragmentation in the long run.
2074 3. Other bins are scanned in increasing size order, using a
2075 chunk big enough to fulfill the request, and splitting off
2076 any remainder. This search is strictly by best-fit; i.e.,
2077 the smallest (with ties going to approximately the least
2078 recently used) chunk that fits is selected.
2080 4. If large enough, the chunk bordering the end of memory
2081 (`top') is split off. (This use of `top' is in accord with
2082 the best-fit search rule. In effect, `top' is treated as
2083 larger (and thus less well fitting) than any other available
2084 chunk since it can be extended to be as large as necessary
2085 (up to system limitations).
2087 5. If the request size meets the mmap threshold and the
2088 system supports mmap, and there are few enough currently
2089 allocated mmapped regions, and a call to mmap succeeds,
2090 the request is allocated via direct memory mapping.
2092 6. Otherwise, the top of memory is extended by
2093 obtaining more space from the system (normally using sbrk,
2094 but definable to anything else via the MORECORE macro).
2095 Memory is gathered from the system (in system page-sized
2096 units) in a way that allows chunks obtained across different
2097 sbrk calls to be consolidated, but does not require
2098 contiguous memory. Thus, it should be safe to intersperse
2099 mallocs with other sbrk calls.
2102 All allocations are made from the the `lowest' part of any found
2103 chunk. (The implementation invariant is that prev_inuse is
2104 always true of any allocated chunk; i.e., that each allocated
2105 chunk borders either a previously allocated and still in-use chunk,
2106 or the base of its memory arena.)
2111 Void_t* mALLOc(size_t bytes)
2113 Void_t* mALLOc(bytes) size_t bytes;
2116 mchunkptr victim; /* inspected/selected chunk */
2117 INTERNAL_SIZE_T victim_size; /* its size */
2118 int idx; /* index for bin traversal */
2119 mbinptr bin; /* associated bin */
2120 mchunkptr remainder; /* remainder from a split */
2121 long remainder_size; /* its size */
2122 int remainder_index; /* its bin index */
2123 unsigned long block; /* block traverser bit */
2124 int startidx; /* first bin of a traversed block */
2125 mchunkptr fwd; /* misc temp for linking */
2126 mchunkptr bck; /* misc temp for linking */
2127 mbinptr q; /* misc temp */
2131 if ((long)bytes < 0) return 0;
2133 nb = request2size(bytes); /* padded request size; */
2135 /* Check for exact match in a bin */
2137 if (is_small_request(nb)) /* Faster version for small requests */
2139 idx = smallbin_index(nb);
2141 /* No traversal or size check necessary for small bins. */
2146 /* Also scan the next one, since it would have a remainder < MINSIZE */
2154 victim_size = chunksize(victim);
2155 unlink(victim, bck, fwd);
2156 set_inuse_bit_at_offset(victim, victim_size);
2157 check_malloced_chunk(victim, nb);
2158 return chunk2mem(victim);
2161 idx += 2; /* Set for bin scan below. We've already scanned 2 bins. */
2166 idx = bin_index(nb);
2169 for (victim = last(bin); victim != bin; victim = victim->bk)
2171 victim_size = chunksize(victim);
2172 remainder_size = victim_size - nb;
2174 if (remainder_size >= (long)MINSIZE) /* too big */
2176 --idx; /* adjust to rescan below after checking last remainder */
2180 else if (remainder_size >= 0) /* exact fit */
2182 unlink(victim, bck, fwd);
2183 set_inuse_bit_at_offset(victim, victim_size);
2184 check_malloced_chunk(victim, nb);
2185 return chunk2mem(victim);
2193 /* Try to use the last split-off remainder */
2195 if ( (victim = last_remainder->fd) != last_remainder)
2197 victim_size = chunksize(victim);
2198 remainder_size = victim_size - nb;
2200 if (remainder_size >= (long)MINSIZE) /* re-split */
2202 remainder = chunk_at_offset(victim, nb);
2203 set_head(victim, nb | PREV_INUSE);
2204 link_last_remainder(remainder);
2205 set_head(remainder, remainder_size | PREV_INUSE);
2206 set_foot(remainder, remainder_size);
2207 check_malloced_chunk(victim, nb);
2208 return chunk2mem(victim);
2211 clear_last_remainder;
2213 if (remainder_size >= 0) /* exhaust */
2215 set_inuse_bit_at_offset(victim, victim_size);
2216 check_malloced_chunk(victim, nb);
2217 return chunk2mem(victim);
2220 /* Else place in bin */
2222 frontlink(victim, victim_size, remainder_index, bck, fwd);
2226 If there are any possibly nonempty big-enough blocks,
2227 search for best fitting chunk by scanning bins in blockwidth units.
2230 if ( (block = idx2binblock(idx)) <= binblocks)
2233 /* Get to the first marked block */
2235 if ( (block & binblocks) == 0)
2237 /* force to an even block boundary */
2238 idx = (idx & ~(BINBLOCKWIDTH - 1)) + BINBLOCKWIDTH;
2240 while ((block & binblocks) == 0)
2242 idx += BINBLOCKWIDTH;
2247 /* For each possibly nonempty block ... */
2250 startidx = idx; /* (track incomplete blocks) */
2251 q = bin = bin_at(idx);
2253 /* For each bin in this block ... */
2256 /* Find and use first big enough chunk ... */
2258 for (victim = last(bin); victim != bin; victim = victim->bk)
2260 victim_size = chunksize(victim);
2261 remainder_size = victim_size - nb;
2263 if (remainder_size >= (long)MINSIZE) /* split */
2265 remainder = chunk_at_offset(victim, nb);
2266 set_head(victim, nb | PREV_INUSE);
2267 unlink(victim, bck, fwd);
2268 link_last_remainder(remainder);
2269 set_head(remainder, remainder_size | PREV_INUSE);
2270 set_foot(remainder, remainder_size);
2271 check_malloced_chunk(victim, nb);
2272 return chunk2mem(victim);
2275 else if (remainder_size >= 0) /* take */
2277 set_inuse_bit_at_offset(victim, victim_size);
2278 unlink(victim, bck, fwd);
2279 check_malloced_chunk(victim, nb);
2280 return chunk2mem(victim);
2285 bin = next_bin(bin);
2287 } while ((++idx & (BINBLOCKWIDTH - 1)) != 0);
2289 /* Clear out the block bit. */
2291 do /* Possibly backtrack to try to clear a partial block */
2293 if ((startidx & (BINBLOCKWIDTH - 1)) == 0)
2295 binblocks &= ~block;
2300 } while (first(q) == q);
2302 /* Get to the next possibly nonempty block */
2304 if ( (block <<= 1) <= binblocks && (block != 0) )
2306 while ((block & binblocks) == 0)
2308 idx += BINBLOCKWIDTH;
2318 /* Try to use top chunk */
2320 /* Require that there be a remainder, ensuring top always exists */
2321 if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2325 /* If big and would otherwise need to extend, try to use mmap instead */
2326 if ((unsigned long)nb >= (unsigned long)mmap_threshold &&
2327 (victim = mmap_chunk(nb)) != 0)
2328 return chunk2mem(victim);
2332 malloc_extend_top(nb);
2333 if ( (remainder_size = chunksize(top) - nb) < (long)MINSIZE)
2334 return 0; /* propagate failure */
2338 set_head(victim, nb | PREV_INUSE);
2339 top = chunk_at_offset(victim, nb);
2340 set_head(top, remainder_size | PREV_INUSE);
2341 check_malloced_chunk(victim, nb);
2342 return chunk2mem(victim);
2355 1. free(0) has no effect.
2357 2. If the chunk was allocated via mmap, it is release via munmap().
2359 3. If a returned chunk borders the current high end of memory,
2360 it is consolidated into the top, and if the total unused
2361 topmost memory exceeds the trim threshold, malloc_trim is
2364 4. Other chunks are consolidated as they arrive, and
2365 placed in corresponding bins. (This includes the case of
2366 consolidating with the current `last_remainder').
2372 void fREe(Void_t* mem)
2374 void fREe(mem) Void_t* mem;
2377 mchunkptr p; /* chunk corresponding to mem */
2378 INTERNAL_SIZE_T hd; /* its head field */
2379 INTERNAL_SIZE_T sz; /* its size */
2380 int idx; /* its bin index */
2381 mchunkptr next; /* next contiguous chunk */
2382 INTERNAL_SIZE_T nextsz; /* its size */
2383 INTERNAL_SIZE_T prevsz; /* size of previous contiguous chunk */
2384 mchunkptr bck; /* misc temp for linking */
2385 mchunkptr fwd; /* misc temp for linking */
2386 int islr; /* track whether merging with last_remainder */
2388 if (mem == 0) /* free(0) has no effect */
2395 if (hd & IS_MMAPPED) /* release mmapped memory. */
2402 check_inuse_chunk(p);
2404 sz = hd & ~PREV_INUSE;
2405 next = chunk_at_offset(p, sz);
2406 nextsz = chunksize(next);
2408 if (next == top) /* merge with top */
2412 if (!(hd & PREV_INUSE)) /* consolidate backward */
2414 prevsz = p->prev_size;
2415 p = chunk_at_offset(p, -((long) prevsz));
2417 unlink(p, bck, fwd);
2420 set_head(p, sz | PREV_INUSE);
2422 if ((unsigned long)(sz) >= (unsigned long)trim_threshold)
2423 malloc_trim(top_pad);
2427 set_head(next, nextsz); /* clear inuse bit */
2431 if (!(hd & PREV_INUSE)) /* consolidate backward */
2433 prevsz = p->prev_size;
2434 p = chunk_at_offset(p, -((long) prevsz));
2437 if (p->fd == last_remainder) /* keep as last_remainder */
2440 unlink(p, bck, fwd);
2443 if (!(inuse_bit_at_offset(next, nextsz))) /* consolidate forward */
2447 if (!islr && next->fd == last_remainder) /* re-insert last_remainder */
2450 link_last_remainder(p);
2453 unlink(next, bck, fwd);
2457 set_head(p, sz | PREV_INUSE);
2460 frontlink(p, sz, idx, bck, fwd);
2471 Chunks that were obtained via mmap cannot be extended or shrunk
2472 unless HAVE_MREMAP is defined, in which case mremap is used.
2473 Otherwise, if their reallocation is for additional space, they are
2474 copied. If for less, they are just left alone.
2476 Otherwise, if the reallocation is for additional space, and the
2477 chunk can be extended, it is, else a malloc-copy-free sequence is
2478 taken. There are several different ways that a chunk could be
2479 extended. All are tried:
2481 * Extending forward into following adjacent free chunk.
2482 * Shifting backwards, joining preceding adjacent space
2483 * Both shifting backwards and extending forward.
2484 * Extending into newly sbrked space
2486 Unless the #define REALLOC_ZERO_BYTES_FREES is set, realloc with a
2487 size argument of zero (re)allocates a minimum-sized chunk.
2489 If the reallocation is for less space, and the new request is for
2490 a `small' (<512 bytes) size, then the newly unused space is lopped
2493 The old unix realloc convention of allowing the last-free'd chunk
2494 to be used as an argument to realloc is no longer supported.
2495 I don't know of any programs still relying on this feature,
2496 and allowing it would also allow too many other incorrect
2497 usages of realloc to be sensible.
2504 Void_t* rEALLOc(Void_t* oldmem, size_t bytes)
2506 Void_t* rEALLOc(oldmem, bytes) Void_t* oldmem; size_t bytes;
2509 INTERNAL_SIZE_T nb; /* padded request size */
2511 mchunkptr oldp; /* chunk corresponding to oldmem */
2512 INTERNAL_SIZE_T oldsize; /* its size */
2514 mchunkptr newp; /* chunk to return */
2515 INTERNAL_SIZE_T newsize; /* its size */
2516 Void_t* newmem; /* corresponding user mem */
2518 mchunkptr next; /* next contiguous chunk after oldp */
2519 INTERNAL_SIZE_T nextsize; /* its size */
2521 mchunkptr prev; /* previous contiguous chunk before oldp */
2522 INTERNAL_SIZE_T prevsize; /* its size */
2524 mchunkptr remainder; /* holds split off extra space from newp */
2525 INTERNAL_SIZE_T remainder_size; /* its size */
2527 mchunkptr bck; /* misc temp for linking */
2528 mchunkptr fwd; /* misc temp for linking */
2530 #ifdef REALLOC_ZERO_BYTES_FREES
2531 if (bytes == 0) { fREe(oldmem); return 0; }
2534 if ((long)bytes < 0) return 0;
2536 /* realloc of null is supposed to be same as malloc */
2537 if (oldmem == 0) return mALLOc(bytes);
2539 newp = oldp = mem2chunk(oldmem);
2540 newsize = oldsize = chunksize(oldp);
2543 nb = request2size(bytes);
2546 if (chunk_is_mmapped(oldp))
2549 newp = mremap_chunk(oldp, nb);
2550 if(newp) return chunk2mem(newp);
2552 /* Note the extra SIZE_SZ overhead. */
2553 if(oldsize - SIZE_SZ >= nb) return oldmem; /* do nothing */
2554 /* Must alloc, copy, free. */
2555 newmem = mALLOc(bytes);
2556 if (newmem == 0) return 0; /* propagate failure */
2557 MALLOC_COPY(newmem, oldmem, oldsize - 2*SIZE_SZ);
2563 check_inuse_chunk(oldp);
2565 if ((long)(oldsize) < (long)(nb))
2568 /* Try expanding forward */
2570 next = chunk_at_offset(oldp, oldsize);
2571 if (next == top || !inuse(next))
2573 nextsize = chunksize(next);
2575 /* Forward into top only if a remainder */
2578 if ((long)(nextsize + newsize) >= (long)(nb + MINSIZE))
2580 newsize += nextsize;
2581 top = chunk_at_offset(oldp, nb);
2582 set_head(top, (newsize - nb) | PREV_INUSE);
2583 set_head_size(oldp, nb);
2584 return chunk2mem(oldp);
2588 /* Forward into next chunk */
2589 else if (((long)(nextsize + newsize) >= (long)(nb)))
2591 unlink(next, bck, fwd);
2592 newsize += nextsize;
2602 /* Try shifting backwards. */
2604 if (!prev_inuse(oldp))
2606 prev = prev_chunk(oldp);
2607 prevsize = chunksize(prev);
2609 /* try forward + backward first to save a later consolidation */
2616 if ((long)(nextsize + prevsize + newsize) >= (long)(nb + MINSIZE))
2618 unlink(prev, bck, fwd);
2620 newsize += prevsize + nextsize;
2621 newmem = chunk2mem(newp);
2622 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2623 top = chunk_at_offset(newp, nb);
2624 set_head(top, (newsize - nb) | PREV_INUSE);
2625 set_head_size(newp, nb);
2630 /* into next chunk */
2631 else if (((long)(nextsize + prevsize + newsize) >= (long)(nb)))
2633 unlink(next, bck, fwd);
2634 unlink(prev, bck, fwd);
2636 newsize += nextsize + prevsize;
2637 newmem = chunk2mem(newp);
2638 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2644 if (prev != 0 && (long)(prevsize + newsize) >= (long)nb)
2646 unlink(prev, bck, fwd);
2648 newsize += prevsize;
2649 newmem = chunk2mem(newp);
2650 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2657 newmem = mALLOc (bytes);
2659 if (newmem == 0) /* propagate failure */
2662 /* Avoid copy if newp is next chunk after oldp. */
2663 /* (This can only happen when new chunk is sbrk'ed.) */
2665 if ( (newp = mem2chunk(newmem)) == next_chunk(oldp))
2667 newsize += chunksize(newp);
2672 /* Otherwise copy, free, and exit */
2673 MALLOC_COPY(newmem, oldmem, oldsize - SIZE_SZ);
2679 split: /* split off extra room in old or expanded chunk */
2681 if (newsize - nb >= MINSIZE) /* split off remainder */
2683 remainder = chunk_at_offset(newp, nb);
2684 remainder_size = newsize - nb;
2685 set_head_size(newp, nb);
2686 set_head(remainder, remainder_size | PREV_INUSE);
2687 set_inuse_bit_at_offset(remainder, remainder_size);
2688 fREe(chunk2mem(remainder)); /* let free() deal with it */
2692 set_head_size(newp, newsize);
2693 set_inuse_bit_at_offset(newp, newsize);
2696 check_inuse_chunk(newp);
2697 return chunk2mem(newp);
2707 memalign requests more than enough space from malloc, finds a spot
2708 within that chunk that meets the alignment request, and then
2709 possibly frees the leading and trailing space.
2711 The alignment argument must be a power of two. This property is not
2712 checked by memalign, so misuse may result in random runtime errors.
2714 8-byte alignment is guaranteed by normal malloc calls, so don't
2715 bother calling memalign with an argument of 8 or less.
2717 Overreliance on memalign is a sure way to fragment space.
2723 Void_t* mEMALIGn(size_t alignment, size_t bytes)
2725 Void_t* mEMALIGn(alignment, bytes) size_t alignment; size_t bytes;
2728 INTERNAL_SIZE_T nb; /* padded request size */
2729 char* m; /* memory returned by malloc call */
2730 mchunkptr p; /* corresponding chunk */
2731 char* brk; /* alignment point within p */
2732 mchunkptr newp; /* chunk to return */
2733 INTERNAL_SIZE_T newsize; /* its size */
2734 INTERNAL_SIZE_T leadsize; /* leading space befor alignment point */
2735 mchunkptr remainder; /* spare room at end to split off */
2736 long remainder_size; /* its size */
2738 if ((long)bytes < 0) return 0;
2740 /* If need less alignment than we give anyway, just relay to malloc */
2742 if (alignment <= MALLOC_ALIGNMENT) return mALLOc(bytes);
2744 /* Otherwise, ensure that it is at least a minimum chunk size */
2746 if (alignment < MINSIZE) alignment = MINSIZE;
2748 /* Call malloc with worst case padding to hit alignment. */
2750 nb = request2size(bytes);
2751 m = (char*)(mALLOc(nb + alignment + MINSIZE));
2753 if (m == 0) return 0; /* propagate failure */
2757 if ((((unsigned long)(m)) % alignment) == 0) /* aligned */
2760 if(chunk_is_mmapped(p))
2761 return chunk2mem(p); /* nothing more to do */
2764 else /* misaligned */
2767 Find an aligned spot inside chunk.
2768 Since we need to give back leading space in a chunk of at
2769 least MINSIZE, if the first calculation places us at
2770 a spot with less than MINSIZE leader, we can move to the
2771 next aligned spot -- we've allocated enough total room so that
2772 this is always possible.
2775 brk = (char*)mem2chunk(((unsigned long)(m + alignment - 1)) & -((signed) alignment));
2776 if ((long)(brk - (char*)(p)) < MINSIZE) brk = brk + alignment;
2778 newp = (mchunkptr)brk;
2779 leadsize = brk - (char*)(p);
2780 newsize = chunksize(p) - leadsize;
2783 if(chunk_is_mmapped(p))
2785 newp->prev_size = p->prev_size + leadsize;
2786 set_head(newp, newsize|IS_MMAPPED);
2787 return chunk2mem(newp);
2791 /* give back leader, use the rest */
2793 set_head(newp, newsize | PREV_INUSE);
2794 set_inuse_bit_at_offset(newp, newsize);
2795 set_head_size(p, leadsize);
2799 assert (newsize >= nb && (((unsigned long)(chunk2mem(p))) % alignment) == 0);
2802 /* Also give back spare room at the end */
2804 remainder_size = chunksize(p) - nb;
2806 if (remainder_size >= (long)MINSIZE)
2808 remainder = chunk_at_offset(p, nb);
2809 set_head(remainder, remainder_size | PREV_INUSE);
2810 set_head_size(p, nb);
2811 fREe(chunk2mem(remainder));
2814 check_inuse_chunk(p);
2815 return chunk2mem(p);
2823 valloc just invokes memalign with alignment argument equal
2824 to the page size of the system (or as near to this as can
2825 be figured out from all the includes/defines above.)
2829 Void_t* vALLOc(size_t bytes)
2831 Void_t* vALLOc(bytes) size_t bytes;
2834 return mEMALIGn (malloc_getpagesize, bytes);
2838 pvalloc just invokes valloc for the nearest pagesize
2839 that will accommodate request
2844 Void_t* pvALLOc(size_t bytes)
2846 Void_t* pvALLOc(bytes) size_t bytes;
2849 size_t pagesize = malloc_getpagesize;
2850 return mEMALIGn (pagesize, (bytes + pagesize - 1) & ~(pagesize - 1));
2855 calloc calls malloc, then zeroes out the allocated chunk.
2860 Void_t* cALLOc(size_t n, size_t elem_size)
2862 Void_t* cALLOc(n, elem_size) size_t n; size_t elem_size;
2866 INTERNAL_SIZE_T csz;
2868 INTERNAL_SIZE_T sz = n * elem_size;
2871 /* check if expand_top called, in which case don't need to clear */
2873 mchunkptr oldtop = top;
2874 INTERNAL_SIZE_T oldtopsize = chunksize(top);
2876 Void_t* mem = mALLOc (sz);
2878 if ((long)n < 0) return 0;
2886 /* Two optional cases in which clearing not necessary */
2890 if (chunk_is_mmapped(p)) return mem;
2896 if (p == oldtop && csz > oldtopsize)
2898 /* clear only the bytes from non-freshly-sbrked memory */
2903 MALLOC_ZERO(mem, csz - SIZE_SZ);
2910 cfree just calls free. It is needed/defined on some systems
2911 that pair it with calloc, presumably for odd historical reasons.
2915 #if !defined(INTERNAL_LINUX_C_LIB) || !defined(__ELF__)
2917 void cfree(Void_t *mem)
2919 void cfree(mem) Void_t *mem;
2930 Malloc_trim gives memory back to the system (via negative
2931 arguments to sbrk) if there is unused memory at the `high' end of
2932 the malloc pool. You can call this after freeing large blocks of
2933 memory to potentially reduce the system-level memory requirements
2934 of a program. However, it cannot guarantee to reduce memory. Under
2935 some allocation patterns, some large free blocks of memory will be
2936 locked between two used chunks, so they cannot be given back to
2939 The `pad' argument to malloc_trim represents the amount of free
2940 trailing space to leave untrimmed. If this argument is zero,
2941 only the minimum amount of memory to maintain internal data
2942 structures will be left (one page or less). Non-zero arguments
2943 can be supplied to maintain enough trailing space to service
2944 future expected allocations without having to re-obtain memory
2947 Malloc_trim returns 1 if it actually released any memory, else 0.
2952 int malloc_trim(size_t pad)
2954 int malloc_trim(pad) size_t pad;
2957 long top_size; /* Amount of top-most memory */
2958 long extra; /* Amount to release */
2959 char* current_brk; /* address returned by pre-check sbrk call */
2960 char* new_brk; /* address returned by negative sbrk call */
2962 unsigned long pagesz = malloc_getpagesize;
2964 top_size = chunksize(top);
2965 extra = ((top_size - pad - MINSIZE + (pagesz-1)) / pagesz - 1) * pagesz;
2967 if (extra < (long)pagesz) /* Not enough memory to release */
2972 /* Test to make sure no one else called sbrk */
2973 current_brk = (char*)(MORECORE (0));
2974 if (current_brk != (char*)(top) + top_size)
2975 return 0; /* Apparently we don't own memory; must fail */
2979 new_brk = (char*)(MORECORE (-extra));
2981 if (new_brk == (char*)(MORECORE_FAILURE)) /* sbrk failed? */
2983 /* Try to figure out what we have */
2984 current_brk = (char*)(MORECORE (0));
2985 top_size = current_brk - (char*)top;
2986 if (top_size >= (long)MINSIZE) /* if not, we are very very dead! */
2988 sbrked_mem = current_brk - sbrk_base;
2989 set_head(top, top_size | PREV_INUSE);
2997 /* Success. Adjust top accordingly. */
2998 set_head(top, (top_size - extra) | PREV_INUSE);
2999 sbrked_mem -= extra;
3012 This routine tells you how many bytes you can actually use in an
3013 allocated chunk, which may be more than you requested (although
3014 often not). You can use this many bytes without worrying about
3015 overwriting other allocated objects. Not a particularly great
3016 programming practice, but still sometimes useful.
3021 size_t malloc_usable_size(Void_t* mem)
3023 size_t malloc_usable_size(mem) Void_t* mem;
3032 if(!chunk_is_mmapped(p))
3034 if (!inuse(p)) return 0;
3035 check_inuse_chunk(p);
3036 return chunksize(p) - SIZE_SZ;
3038 return chunksize(p) - 2*SIZE_SZ;
3045 /* Utility to update current_mallinfo for malloc_stats and mallinfo() */
3047 static void malloc_update_mallinfo()
3056 INTERNAL_SIZE_T avail = chunksize(top);
3057 int navail = ((long)(avail) >= (long)MINSIZE)? 1 : 0;
3059 for (i = 1; i < NAV; ++i)
3062 for (p = last(b); p != b; p = p->bk)
3065 check_free_chunk(p);
3066 for (q = next_chunk(p);
3067 q < top && inuse(q) && (long)(chunksize(q)) >= (long)MINSIZE;
3069 check_inuse_chunk(q);
3071 avail += chunksize(p);
3076 current_mallinfo.ordblks = navail;
3077 current_mallinfo.uordblks = sbrked_mem - avail;
3078 current_mallinfo.fordblks = avail;
3079 current_mallinfo.hblks = n_mmaps;
3080 current_mallinfo.hblkhd = mmapped_mem;
3081 current_mallinfo.keepcost = chunksize(top);
3091 Prints on stderr the amount of space obtain from the system (both
3092 via sbrk and mmap), the maximum amount (which may be more than
3093 current if malloc_trim and/or munmap got called), the maximum
3094 number of simultaneous mmap regions used, and the current number
3095 of bytes allocated via malloc (or realloc, etc) but not yet
3096 freed. (Note that this is the number of bytes allocated, not the
3097 number requested. It will be larger than the number requested
3098 because of alignment and bookkeeping overhead.)
3104 malloc_update_mallinfo();
3105 fprintf(stderr, "max system bytes = %10u\n",
3106 (unsigned int)(max_total_mem));
3107 fprintf(stderr, "system bytes = %10u\n",
3108 (unsigned int)(sbrked_mem + mmapped_mem));
3109 fprintf(stderr, "in use bytes = %10u\n",
3110 (unsigned int)(current_mallinfo.uordblks + mmapped_mem));
3112 fprintf(stderr, "max mmap regions = %10u\n",
3113 (unsigned int)max_n_mmaps);
3118 mallinfo returns a copy of updated current mallinfo.
3121 struct mallinfo mALLINFo()
3123 malloc_update_mallinfo();
3124 return current_mallinfo;
3133 mallopt is the general SVID/XPG interface to tunable parameters.
3134 The format is to provide a (parameter-number, parameter-value) pair.
3135 mallopt then sets the corresponding parameter to the argument
3136 value if it can (i.e., so long as the value is meaningful),
3137 and returns 1 if successful else 0.
3139 See descriptions of tunable parameters above.
3144 int mALLOPt(int param_number, int value)
3146 int mALLOPt(param_number, value) int param_number; int value;
3149 switch(param_number)
3151 case M_TRIM_THRESHOLD:
3152 trim_threshold = value; return 1;
3154 top_pad = value; return 1;
3155 case M_MMAP_THRESHOLD:
3156 mmap_threshold = value; return 1;
3159 n_mmaps_max = value; return 1;
3161 if (value != 0) return 0; else n_mmaps_max = value; return 1;
3173 V2.6.6 Sun Dec 5 07:42:19 1999 Doug Lea (dl at gee)
3174 * return null for negative arguments
3175 * Added Several WIN32 cleanups from Martin C. Fong <mcfong@yahoo.com>
3176 * Add 'LACKS_SYS_PARAM_H' for those systems without 'sys/param.h'
3177 (e.g. WIN32 platforms)
3178 * Cleanup up header file inclusion for WIN32 platforms
3179 * Cleanup code to avoid Microsoft Visual C++ compiler complaints
3180 * Add 'USE_DL_PREFIX' to quickly allow co-existence with existing
3181 memory allocation routines
3182 * Set 'malloc_getpagesize' for WIN32 platforms (needs more work)
3183 * Use 'assert' rather than 'ASSERT' in WIN32 code to conform to
3184 usage of 'assert' in non-WIN32 code
3185 * Improve WIN32 'sbrk()' emulation's 'findRegion()' routine to
3187 * Always call 'fREe()' rather than 'free()'
3189 V2.6.5 Wed Jun 17 15:57:31 1998 Doug Lea (dl at gee)
3190 * Fixed ordering problem with boundary-stamping
3192 V2.6.3 Sun May 19 08:17:58 1996 Doug Lea (dl at gee)
3193 * Added pvalloc, as recommended by H.J. Liu
3194 * Added 64bit pointer support mainly from Wolfram Gloger
3195 * Added anonymously donated WIN32 sbrk emulation
3196 * Malloc, calloc, getpagesize: add optimizations from Raymond Nijssen
3197 * malloc_extend_top: fix mask error that caused wastage after
3199 * Add linux mremap support code from HJ Liu
3201 V2.6.2 Tue Dec 5 06:52:55 1995 Doug Lea (dl at gee)
3202 * Integrated most documentation with the code.
3203 * Add support for mmap, with help from
3204 Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3205 * Use last_remainder in more cases.
3206 * Pack bins using idea from colin@nyx10.cs.du.edu
3207 * Use ordered bins instead of best-fit threshhold
3208 * Eliminate block-local decls to simplify tracing and debugging.
3209 * Support another case of realloc via move into top
3210 * Fix error occuring when initial sbrk_base not word-aligned.
3211 * Rely on page size for units instead of SBRK_UNIT to
3212 avoid surprises about sbrk alignment conventions.
3213 * Add mallinfo, mallopt. Thanks to Raymond Nijssen
3214 (raymond@es.ele.tue.nl) for the suggestion.
3215 * Add `pad' argument to malloc_trim and top_pad mallopt parameter.
3216 * More precautions for cases where other routines call sbrk,
3217 courtesy of Wolfram Gloger (Gloger@lrz.uni-muenchen.de).
3218 * Added macros etc., allowing use in linux libc from
3219 H.J. Lu (hjl@gnu.ai.mit.edu)
3220 * Inverted this history list
3222 V2.6.1 Sat Dec 2 14:10:57 1995 Doug Lea (dl at gee)
3223 * Re-tuned and fixed to behave more nicely with V2.6.0 changes.
3224 * Removed all preallocation code since under current scheme
3225 the work required to undo bad preallocations exceeds
3226 the work saved in good cases for most test programs.
3227 * No longer use return list or unconsolidated bins since
3228 no scheme using them consistently outperforms those that don't
3229 given above changes.
3230 * Use best fit for very large chunks to prevent some worst-cases.
3231 * Added some support for debugging
3233 V2.6.0 Sat Nov 4 07:05:23 1995 Doug Lea (dl at gee)
3234 * Removed footers when chunks are in use. Thanks to
3235 Paul Wilson (wilson@cs.texas.edu) for the suggestion.
3237 V2.5.4 Wed Nov 1 07:54:51 1995 Doug Lea (dl at gee)
3238 * Added malloc_trim, with help from Wolfram Gloger
3239 (wmglo@Dent.MED.Uni-Muenchen.DE).
3241 V2.5.3 Tue Apr 26 10:16:01 1994 Doug Lea (dl at g)
3243 V2.5.2 Tue Apr 5 16:20:40 1994 Doug Lea (dl at g)
3244 * realloc: try to expand in both directions
3245 * malloc: swap order of clean-bin strategy;
3246 * realloc: only conditionally expand backwards
3247 * Try not to scavenge used bins
3248 * Use bin counts as a guide to preallocation
3249 * Occasionally bin return list chunks in first scan
3250 * Add a few optimizations from colin@nyx10.cs.du.edu
3252 V2.5.1 Sat Aug 14 15:40:43 1993 Doug Lea (dl at g)
3253 * faster bin computation & slightly different binning
3254 * merged all consolidations to one part of malloc proper
3255 (eliminating old malloc_find_space & malloc_clean_bin)
3256 * Scan 2 returns chunks (not just 1)
3257 * Propagate failure in realloc if malloc returns 0
3258 * Add stuff to allow compilation on non-ANSI compilers
3259 from kpv@research.att.com
3261 V2.5 Sat Aug 7 07:41:59 1993 Doug Lea (dl at g.oswego.edu)
3262 * removed potential for odd address access in prev_chunk
3263 * removed dependency on getpagesize.h
3264 * misc cosmetics and a bit more internal documentation
3265 * anticosmetics: mangled names in macros to evade debugger strangeness
3266 * tested on sparc, hp-700, dec-mips, rs6000
3267 with gcc & native cc (hp, dec only) allowing
3268 Detlefs & Zorn comparison study (in SIGPLAN Notices.)
3270 Trial version Fri Aug 28 13:14:29 1992 Doug Lea (dl at g.oswego.edu)
3271 * Based loosely on libg++-1.2X malloc. (It retains some of the overall
3272 structure of old version, but most details differ.)