pell@isy.liu.se (P{r Emanuelsson) (08/12/89)
As you might have noticed, the GNU malloc is not friendly towards
allocation of chunks an even power-of-two in length. I think this is a bit
sad, since power-of-two allocation is often used for large allocations,
where this behaviour hurts the most. In our image processing environment I
was only able to allocate half as many images when switching to the GNU
malloc.
I fixed this. The optimal size for allocation is now any power-of-two.
This works only because this malloc never tries to split a large
free block when there are no smaller ones left. If this feature is
added to GNU malloc, then my fixes won't work.
I'd like you to try it out and report possible bugs to me or this list.
The former probably most appropriate, since this stuff is not approved
by the FSF.
To enable my fixes, add the -DNO_ALIGN flag when compiling. With NO_ALIGN
undefined, this should work exactly as before (I hope).
I've used this as the sole memory allocator for a large G++-SunView project
a couple of days with no problems at all. But no warranty!
The fixes are minor, but since the source is small I'll post it here.
Saves people the trouble of digging it out of the other GNU applications.
/Pell
--
"Don't think; let the machine do it for you!"
-- E. C. Berkeley
Dept. of Electrical Engineering pell@isy.liu.se
University of Linkoping, Sweden ...!uunet!isy.liu.se!pell
--
/* dynamic memory allocation for GNU.
Copyright (C) 1985, 1987 Free Software Foundation, Inc.
NO WARRANTY
BECAUSE THIS PROGRAM IS LICENSED FREE OF CHARGE, WE PROVIDE ABSOLUTELY
NO WARRANTY, TO THE EXTENT PERMITTED BY APPLICABLE STATE LAW. EXCEPT
WHEN OTHERWISE STATED IN WRITING, FREE SOFTWARE FOUNDATION, INC,
RICHARD M. STALLMAN AND/OR OTHER PARTIES PROVIDE THIS PROGRAM "AS IS"
WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING,
BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE QUALITY
AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE PROGRAM PROVE
DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR
CORRECTION.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW WILL RICHARD M.
STALLMAN, THE FREE SOFTWARE FOUNDATION, INC., AND/OR ANY OTHER PARTY
WHO MAY MODIFY AND REDISTRIBUTE THIS PROGRAM AS PERMITTED BELOW, BE
LIABLE TO YOU FOR DAMAGES, INCLUDING ANY LOST PROFITS, LOST MONIES, OR
OTHER SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
USE OR INABILITY TO USE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR
DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY THIRD PARTIES OR
A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS) THIS
PROGRAM, EVEN IF YOU HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH
DAMAGES, OR FOR ANY CLAIM BY ANY OTHER PARTY.
GENERAL PUBLIC LICENSE TO COPY
1. You may copy and distribute verbatim copies of this source file
as you receive it, in any medium, provided that you conspicuously and
appropriately publish on each copy a valid copyright notice "Copyright
(C) 1985 Free Software Foundation, Inc."; and include following the
copyright notice a verbatim copy of the above disclaimer of warranty
and of this License. You may charge a distribution fee for the
physical act of transferring a copy.
2. You may modify your copy or copies of this source file or
any portion of it, and copy and distribute such modifications under
the terms of Paragraph 1 above, provided that you also do the following:
a) cause the modified files to carry prominent notices stating
that you changed the files and the date of any change; and
b) cause the whole of any work that you distribute or publish,
that in whole or in part contains or is a derivative of this
program or any part thereof, to be licensed at no charge to all
third parties on terms identical to those contained in this
License Agreement (except that you may choose to grant more extensive
warranty protection to some or all third parties, at your option).
c) You may charge a distribution fee for the physical act of
transferring a copy, and you may at your option offer warranty
protection in exchange for a fee.
Mere aggregation of another unrelated program with this program (or its
derivative) on a volume of a storage or distribution medium does not bring
the other program under the scope of these terms.
3. You may copy and distribute this program (or a portion or derivative
of it, under Paragraph 2) in object code or executable form under the terms
of Paragraphs 1 and 2 above provided that you also do one of the following:
a) accompany it with the complete corresponding machine-readable
source code, which must be distributed under the terms of
Paragraphs 1 and 2 above; or,
b) accompany it with a written offer, valid for at least three
years, to give any third party free (except for a nominal
shipping charge) a complete machine-readable copy of the
corresponding source code, to be distributed under the terms of
Paragraphs 1 and 2 above; or,
c) accompany it with the information you received as to where the
corresponding source code may be obtained. (This alternative is
allowed only for noncommercial distribution and only if you
received the program in object code or executable form alone.)
For an executable file, complete source code means all the source code for
all modules it contains; but, as a special exception, it need not include
source code for modules which are standard libraries that accompany the
operating system on which the executable file runs.
4. You may not copy, sublicense, distribute or transfer this program
except as expressly provided under this License Agreement. Any attempt
otherwise to copy, sublicense, distribute or transfer this program is void and
your rights to use the program under this License agreement shall be
automatically terminated. However, parties who have received computer
software programs from you with this License Agreement will not have
their licenses terminated so long as such parties remain in full compliance.
5. If you wish to incorporate parts of this program into other free
programs whose distribution conditions are different, write to the Free
Software Foundation at 675 Mass Ave, Cambridge, MA 02139. We have not yet
worked out a simple rule that can be stated here, but we will often permit
this. We will be guided by the two goals of preserving the free status of
all derivatives of our free software and of promoting the sharing and reuse of
software.
In other words, you are welcome to use, share and improve this program.
You are forbidden to forbid anyone else to use, share and improve
what you give them. Help stamp out software-hoarding! */
/*
* @(#)nmalloc.c 1 (Caltech) 2/21/82
*
* U of M Modified: 20 Jun 1983 ACT: strange hacks for Emacs
*
* Nov 1983, Mike@BRL, Added support for 4.1C/4.2 BSD.
*
* This is a very fast storage allocator. It allocates blocks of a small
* number of different sizes, and keeps free lists of each size. Blocks
* that don't exactly fit are passed up to the next larger size. In this
* implementation, the available sizes are (2^n)-4 (or -16) bytes long.
* This is designed for use in a program that uses vast quantities of
* memory, but bombs when it runs out. To make it a little better, it
* warns the user when he starts to get near the end.
*
* June 84, ACT: modified rcheck code to check the range given to malloc,
* rather than the range determined by the 2-power used.
*
* Jan 85, RMS: calls malloc_warning to issue warning on nearly full.
* No longer Emacs-specific; can serve as all-purpose malloc for GNU.
* You should call malloc_init to reinitialize after loading dumped Emacs.
* Call malloc_stats to get info on memory stats if MSTATS turned on.
* realloc knows how to return same block given, just changing its size,
* if the power of 2 is correct.
*
* Aug 89, Pell: Added NO_ALIGN flag. This tells malloc that we don't want
* it to align sbrk() calls, presumably because it gains nothing on this
* particular OS. Defining NO_ALIGN also has the effect that the optimal
* allocation size is power-of-two. Thus, if you allocate large amounts
* of power-of-two blocks, you want to define NO_ALIGN. Without NO_ALIGN,
* malloc() will allocate twice as much as you ask for.
* NO_ALIGN works fine on SunOS 4 at least.
*/
/*
* nextf[i] is the pointer to the next free block of size 2^(i+3). The
* smallest allocatable block is 8 bytes. The overhead information will
* go in the first int of the block, and the returned pointer will point
* to the second.
*
#ifdef MSTATS
* nmalloc[i] is the difference between the number of mallocs and frees
* for a given block size.
#endif MSTATS
*/
#ifdef emacs
/* config.h specifies which kind of system this is. */
#include "config.h"
#else
/* Determine which kind of system this is. */
#include <signal.h>
#ifndef SIGTSTP
#ifndef VMS
#ifndef USG
#define USG
#endif
#endif /* not VMS */
#else /* SIGTSTP */
#ifdef SIGIO
#define BSD4_2
#endif /* SIGIO */
#endif /* SIGTSTP */
#endif /* not emacs */
/* Define getpagesize () if the system does not. */
#include "getpagesize.h"
#ifndef BSD4_2
#ifndef USG
#include <sys/vlimit.h> /* warn the user when near the end */
#endif /* not USG */
#else /* if BSD4_2 */
#include <sys/time.h>
#include <sys/resource.h>
#endif /* BSD4_2 */
extern char *start_of_data ();
#ifdef BSD
#ifndef DATA_SEG_BITS
#define start_of_data() &etext
#endif
#endif
#ifndef emacs
#define start_of_data() &etext
#endif
#define ISALLOC ((char) 0xf7) /* magic byte that implies allocation */
#define ISFREE ((char) 0x54) /* magic byte that implies free block */
/* this is for error checking only */
#define ISMEMALIGN ((char) 0xd6) /* Stored before the value returned by
memalign, with the rest of the word
being the distance to the true
beginning of the block. */
extern char etext;
/* These two are for user programs to look at, when they are interested. */
unsigned int malloc_sbrk_used; /* amount of data space used now */
unsigned int malloc_sbrk_unused; /* amount more we can have */
/* start of data space; can be changed by calling init_malloc */
static char *data_space_start;
#ifdef MSTATS
static int nmalloc[30];
static int nmal, nfre;
#endif /* MSTATS */
/* If range checking is not turned on, all we have is a flag indicating
whether memory is allocated, an index in nextf[], and a size field; to
realloc() memory we copy either size bytes or 1<<(index+3) bytes depending
on whether the former can hold the exact size (given the value of
'index'). If range checking is on, we always need to know how much space
is allocated, so the 'size' field is never used. */
struct mhead {
char mh_alloc; /* ISALLOC or ISFREE */
char mh_index; /* index in nextf[] */
/* Remainder are valid only when block is allocated */
unsigned short mh_size; /* size, if < 0x10000 */
#ifdef rcheck
unsigned mh_nbytes; /* number of bytes allocated */
int mh_magic4; /* should be == MAGIC4 */
#endif /* rcheck */
};
/* Access free-list pointer of a block.
It is stored at block + 4.
This is not a field in the mhead structure
because we want sizeof (struct mhead)
to describe the overhead for when the block is in use,
and we do not want the free-list pointer to count in that. */
#define CHAIN(a) \
(*(struct mhead **) (sizeof (char *) + (char *) (a)))
#ifdef rcheck
/* To implement range checking, we write magic values in at the beginning and
end of each allocated block, and make sure they are undisturbed whenever a
free or a realloc occurs. */
/* Written in each of the 4 bytes following the block's real space */
#define MAGIC1 0x55
/* Written in the 4 bytes before the block's real space */
#define MAGIC4 0x55555555
#define ASSERT(p) if (!(p)) botch("p"); else
#define EXTRA 4 /* 4 bytes extra for MAGIC1s */
#else
#define ASSERT(p)
#define EXTRA 0
#endif /* rcheck */
/* nextf[i] is free list of blocks of size 2**(i + 3) */
static struct mhead *nextf[30];
/* busy[i] is nonzero while allocation of block size i is in progress. */
static char busy[30];
/* Number of bytes of writable memory we can expect to be able to get */
static unsigned int lim_data;
/* Level number of warnings already issued.
0 -- no warnings issued.
1 -- 75% warning already issued.
2 -- 85% warning already issued.
*/
static int warnlevel;
/* Function to call to issue a warning;
0 means don't issue them. */
static void (*warnfunction) ();
/* nonzero once initial bunch of free blocks made */
static int gotpool;
char *_malloc_base;
static void getpool ();
/* Cause reinitialization based on job parameters;
also declare where the end of pure storage is. */
void
malloc_init (start, warnfun)
char *start;
void (*warnfun) ();
{
if (start)
data_space_start = start;
lim_data = 0;
warnlevel = 0;
warnfunction = warnfun;
}
/* Return the maximum size to which MEM can be realloc'd
without actually requiring copying. */
int
malloc_usable_size (mem)
char *mem;
{
int blocksize = 8 << (((struct mhead *) mem) - 1) -> mh_index;
#ifdef NO_ALIGN
return blocksize;
#else
return blocksize - sizeof (struct mhead) - EXTRA;
#endif
}
static void
morecore (nu) /* ask system for more memory */
register int nu; /* size index to get more of */
{
char *sbrk ();
register char *cp;
register int nblks;
register unsigned int siz;
int oldmask;
int extra=0, eblks=0;
#ifdef BSD
#ifndef BSD4_1
/* ?? There was a suggestion not to block SIGILL, somehow for GDB's sake. */
oldmask = sigsetmask (-1);
#endif
#endif
if (!data_space_start)
{
data_space_start = start_of_data ();
}
if (lim_data == 0)
get_lim_data ();
/* On initial startup, get two blocks of each size up to 1k bytes */
if (!gotpool)
{ getpool (); getpool (); gotpool = 1; }
/* Find current end of memory and issue warning if getting near max */
#ifndef VMS
/* Maximum virtual memory on VMS is difficult to calculate since it
* depends on several dynmacially changing things. Also, alignment
* isn't that important. That is why much of the code here is ifdef'ed
* out for VMS systems.
*/
cp = sbrk (0);
siz = cp - data_space_start;
if (warnfunction)
switch (warnlevel)
{
case 0:
if (siz > (lim_data / 4) * 3)
{
warnlevel++;
(*warnfunction) ("Warning: past 75% of memory limit");
}
break;
case 1:
if (siz > (lim_data / 20) * 17)
{
warnlevel++;
(*warnfunction) ("Warning: past 85% of memory limit");
}
break;
case 2:
if (siz > (lim_data / 20) * 19)
{
warnlevel++;
(*warnfunction) ("Warning: past 95% of memory limit");
}
break;
}
#ifdef NO_ALIGN
if ((int) cp & 7) /* land on 8-byte boundaries */
sbrk (8 - ((int) cp & 7));
#else
if ((int) cp & 0x3ff) /* land on 1K boundaries */
sbrk (1024 - ((int) cp & 0x3ff));
#endif
#endif /* not VMS */
/* Take at least 2k, and figure out how many blocks of the desired size
we're about to get */
nblks = 1;
if ((siz = nu) < 8)
nblks = 1 << ((siz = 8) - nu);
#ifdef NO_ALIGN
extra = ((sizeof (struct mhead) + 7) & ~7) + EXTRA;
eblks = extra * nblks;
#endif
if ((cp = sbrk ((1 << (siz + 3)) + eblks)) == (char *) -1)
{
#ifdef BSD
#ifndef BSD4_1
sigsetmask (oldmask);
#endif
#endif
return; /* no more room! */
}
malloc_sbrk_used = siz;
malloc_sbrk_unused = lim_data - siz;
#ifndef VMS
if ((int) cp & 7)
{ /* shouldn't happen, but just in case */
cp = (char *) (((int) cp + 8) & ~7);
nblks--;
}
#endif /* not VMS */
/* save new header and link the nblks blocks together */
nextf[nu] = (struct mhead *) cp;
siz = (1 << (nu + 3)) + extra;
while (1)
{
((struct mhead *) cp) -> mh_alloc = ISFREE;
((struct mhead *) cp) -> mh_index = nu;
if (--nblks <= 0) break;
CHAIN ((struct mhead *) cp) = (struct mhead *) (cp + siz);
cp += siz;
}
CHAIN ((struct mhead *) cp) = 0;
#ifdef BSD
#ifndef BSD4_1
sigsetmask (oldmask);
#endif
#endif
}
static void
getpool ()
{
register int nu;
char * sbrk ();
register char *cp = sbrk (0);
register int extra = 0;
#ifdef NO_ALIGN
if ((int) cp & 7) /* land on 8-byte boundaries */
sbrk (8 - ((int) cp & 7));
#else
if ((int) cp & 0x3ff) /* land on 1K boundaries */
sbrk (1024 - ((int) cp & 0x3ff));
#endif
/* Record address of start of space allocated by malloc. */
if (_malloc_base == 0)
_malloc_base = cp;
#ifdef NO_ALIGN
extra = ((sizeof (struct mhead) + 7) & ~7) + EXTRA;
#endif
/* Get 2k of storage (+extra)*/
cp = sbrk (04000 + 8*extra);
if (cp == (char *) -1)
return;
/* Divide it into an initial 8-word block
plus one block of size 2**nu for nu = 3 ... 9. */
CHAIN (cp) = nextf[0];
nextf[0] = (struct mhead *) cp;
((struct mhead *) cp) -> mh_alloc = ISFREE;
((struct mhead *) cp) -> mh_index = 0;
cp += 8+extra;
for (nu = 0; nu < 7; nu++)
{
CHAIN (cp) = nextf[nu];
nextf[nu] = (struct mhead *) cp;
((struct mhead *) cp) -> mh_alloc = ISFREE;
((struct mhead *) cp) -> mh_index = nu;
cp += (8 << nu) + extra;
}
}
char *
malloc (n) /* get a block */
unsigned n;
{
register struct mhead *p;
register unsigned int nbytes;
register int nunits = 0;
register int extra = 0;
/* Figure out how many bytes are required, rounding up to the nearest
multiple of 8, then figure out which nestf[] area to use.
Both the beginning of the header and the beginning of the
block should be on an eight byte boundary. */
#ifndef NO_ALIGN
extra = ((sizeof *p + 7) & ~7) + EXTRA;
#endif /* else extra=0 */
nbytes = (n + extra + 7) & ~7;
{
register unsigned int shiftr = (nbytes - 1) >> 2;
while (shiftr >>= 1)
nunits++;
}
/* In case this is reentrant use of malloc from signal handler,
pick a block size that no other malloc level is currently
trying to allocate. That's the easiest harmless way not to
interfere with the other level of execution. */
while (busy[nunits]) nunits++;
busy[nunits] = 1;
/* If there are no blocks of the appropriate size, go get some */
/* COULD SPLIT UP A LARGER BLOCK HERE ... ACT */
if (nextf[nunits] == 0)
morecore (nunits);
/* Get one block off the list, and set the new list head */
if ((p = nextf[nunits]) == 0)
{
busy[nunits] = 0;
return 0;
}
nextf[nunits] = CHAIN (p);
busy[nunits] = 0;
/* Check for free block clobbered */
/* If not for this check, we would gobble a clobbered free chain ptr */
/* and bomb out on the NEXT allocate of this size block */
if (p -> mh_alloc != ISFREE || p -> mh_index != nunits)
#ifdef rcheck
botch ("block on free list clobbered");
#else /* not rcheck */
abort ();
#endif /* not rcheck */
/* Fill in the info, and if range checking, set up the magic numbers */
p -> mh_alloc = ISALLOC;
#ifdef rcheck
p -> mh_nbytes = n;
p -> mh_magic4 = MAGIC4;
{
/* Get the location n after the beginning of the user's space. */
register char *m = (char *) p + ((sizeof *p + 7) & ~7) + n;
*m++ = MAGIC1, *m++ = MAGIC1, *m++ = MAGIC1, *m = MAGIC1;
}
#else /* not rcheck */
p -> mh_size = n;
#endif /* not rcheck */
#ifdef MSTATS
nmalloc[nunits]++;
nmal++;
#endif /* MSTATS */
return (char *) p + ((sizeof *p + 7) & ~7);
}
free (mem)
char *mem;
{
register struct mhead *p;
{
register char *ap = mem;
if (ap == 0)
return;
p = (struct mhead *) (ap - ((sizeof *p + 7) & ~7));
if (p -> mh_alloc == ISMEMALIGN)
{
ap -= p->mh_size;
p = (struct mhead *) (ap - ((sizeof *p + 7) & ~7));
}
#ifndef rcheck
if (p -> mh_alloc != ISALLOC)
abort ();
#else rcheck
if (p -> mh_alloc != ISALLOC)
{
if (p -> mh_alloc == ISFREE)
botch ("free: Called with already freed block argument\n");
else
botch ("free: Called with bad argument\n");
}
ASSERT (p -> mh_magic4 == MAGIC4);
ap += p -> mh_nbytes;
ASSERT (*ap++ == MAGIC1); ASSERT (*ap++ == MAGIC1);
ASSERT (*ap++ == MAGIC1); ASSERT (*ap == MAGIC1);
#endif /* rcheck */
}
{
register int nunits = p -> mh_index;
ASSERT (nunits <= 29);
p -> mh_alloc = ISFREE;
/* Protect against signal handlers calling malloc. */
busy[nunits] = 1;
/* Put this block on the free list. */
CHAIN (p) = nextf[nunits];
nextf[nunits] = p;
busy[nunits] = 0;
#ifdef MSTATS
nmalloc[nunits]--;
nfre++;
#endif /* MSTATS */
}
}
char *
realloc (mem, n)
char *mem;
register unsigned n;
{
register struct mhead *p;
register unsigned int tocopy;
register unsigned int nbytes;
register int nunits;
register int extra = 0;
if (mem == 0)
return malloc (n);
p = (struct mhead *) (mem - ((sizeof *p + 7) & ~7));
nunits = p -> mh_index;
ASSERT (p -> mh_alloc == ISALLOC);
#ifdef rcheck
ASSERT (p -> mh_magic4 == MAGIC4);
{
register char *m = mem + (tocopy = p -> mh_nbytes);
ASSERT (*m++ == MAGIC1); ASSERT (*m++ == MAGIC1);
ASSERT (*m++ == MAGIC1); ASSERT (*m == MAGIC1);
}
#else /* not rcheck */
if (p -> mh_index >= 13)
tocopy = (1 << (p -> mh_index + 3)) - ((sizeof *p + 7) & ~7);
else
tocopy = p -> mh_size;
#endif /* not rcheck */
#ifndef NO_ALIGN
extra = ((sizeof *p + 7) & ~7) + EXTRA;
#endif /* else extra=0 */
/* See if desired size rounds to same power of 2 as actual size. */
nbytes = (n + extra + 7) & ~7;
/* If ok, use the same block, just marking its size as changed. */
if (nbytes > (4 << nunits) && nbytes <= (8 << nunits))
{
#ifdef rcheck
register char *m = mem + tocopy;
*m++ = 0; *m++ = 0; *m++ = 0; *m++ = 0;
p-> mh_nbytes = n;
m = mem + n;
*m++ = MAGIC1; *m++ = MAGIC1; *m++ = MAGIC1; *m++ = MAGIC1;
#else /* not rcheck */
p -> mh_size = n;
#endif /* not rcheck */
return mem;
}
if (n < tocopy)
tocopy = n;
{
register char *new;
if ((new = malloc (n)) == 0)
return 0;
bcopy (mem, new, tocopy);
free (mem);
return new;
}
}
#ifndef VMS
char *
memalign (alignment, size)
unsigned alignment, size;
{
register char *ptr = malloc (size + alignment);
register char *aligned;
register struct mhead *p;
if (ptr == 0)
return 0;
/* If entire block has the desired alignment, just accept it. */
if (((int) ptr & (alignment - 1)) == 0)
return ptr;
/* Otherwise, get address of byte in the block that has that alignment. */
aligned = (char *) (((int) ptr + alignment - 1) & -alignment);
/* Store a suitable indication of how to free the block,
so that free can find the true beginning of it. */
p = (struct mhead *) aligned - 1;
p -> mh_size = aligned - ptr;
p -> mh_alloc = ISMEMALIGN;
return aligned;
}
#ifndef hpux
/* This runs into trouble with getpagesize on HPUX.
Patching out seems cleaner than the ugly fix needed. */
char *
valloc (size)
{
return memalign (getpagesize (), size);
}
#endif /* not hpux */
#endif /* not VMS */
#ifdef MSTATS
/* Return statistics describing allocation of blocks of size 2**n. */
struct mstats_value
{
int blocksize;
int nfree;
int nused;
};
struct mstats_value
malloc_stats (size)
int size;
{
struct mstats_value v;
register int i;
register struct mhead *p;
v.nfree = 0;
if (size < 0 || size >= 30)
{
v.blocksize = 0;
v.nused = 0;
return v;
}
v.blocksize = 1 << (size + 3);
v.nused = nmalloc[size];
for (p = nextf[size]; p; p = CHAIN (p))
v.nfree++;
return v;
}
int
malloc_mem_used ()
{
int i;
int size_used;
size_used = 0;
for (i = 0; i < 30; i++)
{
int allocation_size = 1 << (i + 3);
struct mhead *p;
size_used += nmalloc[i] * allocation_size;
}
return size_used;
}
int
malloc_mem_free ()
{
int i;
int size_unused;
size_unused = 0;
for (i = 0; i < 30; i++)
{
int allocation_size = 1 << (i + 3);
struct mhead *p;
for (p = nextf[i]; p ; p = CHAIN (p))
size_unused += allocation_size;
}
return size_unused;
}
#endif /* MSTATS */
/*
* This function returns the total number of bytes that the process
* will be allowed to allocate via the sbrk(2) system call. On
* BSD systems this is the total space allocatable to stack and
* data. On USG systems this is the data space only.
*/
#ifdef USG
get_lim_data ()
{
extern long ulimit ();
#ifdef ULIMIT_BREAK_VALUE
lim_data = ULIMIT_BREAK_VALUE;
#else
lim_data = ulimit (3, 0);
#endif
lim_data -= (long) data_space_start;
}
#else /* not USG */
#ifndef BSD4_2
get_lim_data ()
{
lim_data = vlimit (LIM_DATA, -1);
}
#else /* BSD4_2 */
get_lim_data ()
{
struct rlimit XXrlimit;
getrlimit (RLIMIT_DATA, &XXrlimit);
#ifdef RLIM_INFINITY
lim_data = XXrlimit.rlim_cur & RLIM_INFINITY; /* soft limit */
#else
lim_data = XXrlimit.rlim_cur; /* soft limit */
#endif
}
#endif /* BSD4_2 */
#endif /* not USG */
#ifdef VMS
/* There is a problem when dumping and restoring things on VMS. Calls
* to SBRK don't necessarily result in contiguous allocation. Dumping
* doesn't work when it isn't. Therefore, we make the initial
* allocation contiguous by allocating a big chunk, and do SBRKs from
* there. Once Emacs has dumped there is no reason to continue
* contiguous allocation, malloc doesn't depend on it.
*
* There is a further problem of using brk and sbrk while using VMS C
* run time library routines malloc, calloc, etc. The documentation
* says that this is a no-no, although I'm not sure why this would be
* a problem. In any case, we remove the necessity to call brk and
* sbrk, by calling calloc (to assure zero filled data) rather than
* sbrk.
*
* VMS_ALLOCATION_SIZE is the size of the allocation array. This
* should be larger than the malloc size before dumping. Making this
* too large will result in the startup procedure slowing down since
* it will require more space and time to map it in.
*
* The value for VMS_ALLOCATION_SIZE in the following define was determined
* by running emacs linked (and a large allocation) with the debugger and
* looking to see how much storage was used. The allocation was 201 pages,
* so I rounded it up to a power of two.
*/
#ifndef VMS_ALLOCATION_SIZE
#define VMS_ALLOCATION_SIZE (512*256)
#endif
/* Use VMS RTL definitions */
#undef sbrk
#undef brk
#undef malloc
int vms_out_initial = 0;
char vms_initial_buffer[VMS_ALLOCATION_SIZE];
static char *vms_current_brk = &vms_initial_buffer;
static char *vms_end_brk = &vms_initial_buffer[VMS_ALLOCATION_SIZE-1];
#include <stdio.h>
char *
sys_sbrk (incr)
int incr;
{
char *sbrk(), *temp, *ptr;
if (vms_out_initial)
{
/* out of initial allocation... */
if (!(temp = malloc (incr)))
temp = (char *) -1;
}
else
{
/* otherwise, go out of our area */
ptr = vms_current_brk + incr; /* new current_brk */
if (ptr <= vms_end_brk)
{
temp = vms_current_brk;
vms_current_brk = ptr;
}
else
{
vms_out_initial = 1; /* mark as out of initial allocation */
if (!(temp = malloc (incr)))
temp = (char *) -1;
}
}
return temp;
}
#endif /* VMS */
--
"Don't think; let the machine do it for you!"
-- E. C. Berkeley
Dept. of Electrical Engineering pell@isy.liu.se
University of Linkoping, Sweden ...!uunet!isy.liu.se!pell