tlsf: move to /lib (it's also used by plugins).

Change-Id: I5e37b28c1ce4608d60b036343f280af3311ad490
This commit is contained in:
Thomas Martitz 2012-03-27 16:27:02 +02:00
parent 9f2ee2f21a
commit c56950ea3a
16 changed files with 26 additions and 27 deletions

23
lib/tlsf/COPYING Normal file
View file

@ -0,0 +1,23 @@
LICENSE INFORMATION
TLSF is released as LGPL and GPL. A copy of both licences can be found in this
directoy. For the GPL licence, the following exception applies.
TLSF is free software; you can redistribute it and/or modify it under terms of
the GNU General Public License as published by the Free Software Foundation;
either version 2, or (at your option) any later version. TLSF is distributed
in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the
implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
the GNU General Public License for more details. You should have received a
copy of the GNU General Public License along with TLSF; see file COPYING. If
not, write to the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139,
USA.
As a special exception, including TLSF header files in a file, or linking with
other files objects to produce an executable application, is merely considered
normal use of the library, and does *not* fall under the heading of "derived
work". Therfore does not by itself cause the resulting executable application
to be covered by the GNU General Public License. This exception does not
however invalidate any other reasons why the executable file might be covered
by the GNU Public License.

119
lib/tlsf/Changelog Normal file
View file

@ -0,0 +1,119 @@
Version History
---------------
-v2.4.4 (October 13 2008)
* Corrected minor syntactic bug on statistic gathering code.
Reported by Tim Cussins and P. Mantegazza.
-v2.4.3 (July 30 2008)
* Minor fixes to compile with the greenhills compiler.
Reported by "Kaya, Sinan SEA" <sinan.kaya@siemens.com>
* Small change in the license in order to include TLSF in the RTEMS
project.
-v2.4.2 (May 16 2008) (Herman ten Brugge)
* Memory usage statistics added again, with cleaner and more compacted
code.
-v2.4.1 (April 30 2008)
* Fixed a bug in the tlsf_realloc function: init the pool automatically
on the first call.
Reported by: Alejandro Mery <amery@geeks.cl>
-v2.4 (Feb 19 2008)
* "rtl_*" functions renamed to "tlsf_*".
* Added the add_new_area function to insert new memory areas to an
existing memory pool.
* A single TLSF pool can manage non-contiguous memory areas.
* Support for mmap and sbrk added.
* The init_memory_pool is not longer needed when used on a system
with mmap or sbrk.
* Removed the get_used_size counting.The same functionality can be
implemented outside the TLSF code.
-v2.3.2 (Sep 27 2007)
* Minor cosmetic code improvements.
-v2.3.1 (Jul 30 2007)
* Fixed some minor bugs in the version 2.3. Herman ten Brugge
<hermantenbrugge@home.nl>
-v2.3 (Jul 28 2007) Released a new version with all the contributions
received from Herman ten Brugge <hermantenbrugge@home.nl>
(This is his summary of changes in the TLSF's code):
* Add 64 bit support. It now runs on x86_64 and solaris64.
* I also tested this on vxworks/32 and solaris/32 and i386/32
processors.
* Remove assembly code. I could not measure any performance difference
on my core2 processor. This also makes the code more portable.
* Moved defines/typedefs from tlsf.h to tlsf.c
* Changed MIN_BLOCK_SIZE to sizeof (free_ptr_t) and BHDR_OVERHEAD to
(sizeof (bhdr_t) - MIN_BLOCK_SIZE). This does not change the fact
that the minumum size is still sizeof (bhdr_t).
* Changed all C++ comment style to C style. (// -> /* ... *./)
* Used ls_bit instead of ffs and ms_bit instead of fls. I did this to
avoid confusion with the standard ffs function which returns
different values.
* Created set_bit/clear_bit fuctions because they are not present
on x86_64.
* Added locking support + extra file target.h to show how to use it.
* Added get_used_size function
* Added rtl_realloc and rtl_calloc function
* Implemented realloc clever support.
* Added some test code in the example directory.
-- Thank you very much for your help Herman!
-v2.2.1 (Oct 23 2006)
* Support for ARMv5 implemented by Adam Scislowicz
<proteuskor@gmail.com>. Thank you for your contribution.
- v2.2.0 (Jun 30 2006) Miguel Masmano & Ismael Ripoll.
* Blocks smaller than 128 bytes are stored on a single
segregated list. The already existing bits maps and data
structures are used.
* Minor code speed-up improvements.
* Worst case response time both on malloc and free improved.
* External fragmantation also improved!.
* Segragared lists are AGAIN sorted by LIFO order. Version
2.1b was proven to be no better than 2.1.
- v2.1b: Allocation policy has been always a LIFO Good-Fit, that
is, between several free blocks in the same range, TLSF will
always allocate the most recently released. In this version of
TLSF, we have implemented a FIFO Good-Fit. However,
fragmentation doesn't seems to be altered so is it worth it?.
- v2.1: Realloc and calloc included again in TLSF 2.0.
- v2.0: In this version, TLSF has been programmed from scratch.
Now the allocator is provided as an unique file. Realloc and
calloc are not longer implemented.
- v1.4: Created the section "Version History". Studied real
behaviour of actual applications (regular applications tend
to require small memory blocks (less than 16 bytes) whereas
TLSF is optimised to be used with blocks larger than 16
bytes: Added special lists to deal with blocks smaller than
16 bytes.
- v1.3: Change of concept, now the main TLSF structure is created
inside of the beginning of the block instead of being an
static structure, allowing multiple TLSFs working at the
same time. Now, TLSF uses specific processor instructions to
deal with bitmaps. TLSF sanity functions added to find TLSF
overflows. The TLSF code will not be RTLinux-oriented any
more.
- v1.1 ... v1.2: Many little bugs fixed, code cleaned and splitted
in several files because of cosmetic requirements.
Starting from TLSF v1.1, MaRTE OS
(http://marte.unican.es) uses the TLSF allocator
as its default memory allocator.
- v0.1 ... v1.0: First implementations were created for testing and
research purposes. Basically TLSF is implemented to
be used by RTLinux-GPL (www.rtlinux-gpl.org), so
it is RTLinux-oriented.

163
lib/tlsf/README Normal file
View file

@ -0,0 +1,163 @@
TLSF Memory Storage allocator implementation.
Version 2.4 Feb 2008
Authors: Miguel Masmano, Ismael Ripoll & Alfons Crespo.
Copyright UPVLC, OCERA Consortium.
TLSF is released in the GPL/LGPL licence. The exact terms of the licence
are described in the COPYING file.
This component provides basic memory allocation functions:
malloc and free, as defined in the standard "C" library.
This allocator was designed to provide real-time performance, that is:
1.- Bounded time malloc and free.
2.- Fast response time.
3.- Efficient memory management, that is low fragmentation.
The worst response time for both malloc and free is O(1).
How to use it:
This code is prepared to be used as a stand-alone code that can be
linked with a regular application or it can be compiled to be a Linux
module (which required the BigPhysicalArea patch). Initially the
module was designed to work jointly with RTLinux-GPL, but it can be
used as a stand alone Linux module.
When compiled as a regular linux process the API is:
Initialisation and destruction functions
----------------------------------------
init_memory_pool may be called before any request or release call:
- size_t init_memory_pool(size_t, void *);
- void destroy_memory_pool(void *);
Request and release functions
-----------------------------
As can be seen, there are two functions for each traditional memory
allocation function (malloc, free, realloc, and calloc). One with the
prefix "tlsf_" and the other with the suffix "_ex".
The versions with the prefix "tlsf_" provides the expected behaviour,
that is, allocating/releasing memory from the default memory pool. The
default memory pool is the last pool initialised by the
init_memory_pool function.
On the other hand, the functions with the prefix "_ex" enable the use of several memory pools.
- void *tlsf_malloc(size_t);
- void *malloc_ex(size_t, void *);
- void tlsf_free(void *ptr);
- void free_ex(void *, void *);
- void *tlsf_realloc(void *ptr, size_t size);
- void *realloc_ex(void *, size_t, void *);
- void *tlsf_calloc(size_t nelem, size_t elem_size);
- void *calloc_ex(size_t, size_t, void *);
EXAMPLE OF USE:
char memory_pool[1024*1024];
{
...
init_memory_pool(1024*1024, memory_pool);
...
ptr1=malloc_ex(100, memory_pool);
ptr2=tlsf_malloc(100); // This function will use memory_pool
...
tlsf_free(ptr2);
free_ex(ptr1, memory_pool);
}
Growing the memory pool
-----------------------
Starting from the version 2.4, the function add_new_area adds an
memory area to an existing memory pool.
- size_t add_new_area(void *, size_t, void *);
This feature is pretty useful when an existing memory pool is running
low and we want to add more free memory to it.
EXAMPLE OF USE:
char memory_pool[1024*1024];
char memory_pool2[1024*1024];
{
...
init_memory_pool(1024*1024, memory_pool);
...
ptr[0]=malloc_ex(1024*256 memory_pool);
ptr[1]=malloc_ex(1024*512, memory_pool);
add_new_area(memory_pool2, 1024*1024, memory_pool);
// Now we have an extra free memory area of 1Mb
// The next malloc may not fail
ptr[2]=malloc_ex(1024*512, memory_pool);
...
}
SBRK and MMAP support
---------------------
The version 2.4 can use the functions SBRK and MMAP to _automatically_
growing the memory pool, before running out of memory.
So, when this feature is enabled, unless the operating system were out
of memory, a malloc operation would not fail due to an "out-of-memory"
error.
To enable this support, compile tlsf.c with the FLAGS -DUSE_MMAP=1 or
-DUSE_SBRK=1 depending on whether you want to use "mmap" or "sbrk" or both.
** By default (default Makefile) this feature is enabled.
EXAMPLE OF USE:
gcc -o tlsf.o -O2 -Wall -DUSE_MMAP=1 -DUSE_SBRK=1
---
If the sbrk/mmap support is enabled and we are _only_ going to use one
memory pool, it is not necessary to call init_memory_pool
EXAMPLE OF USE (with MMAP/SBRK support enabled):
{
...
ptr2=tlsf_malloc(100); // This function will use memory_pool
...
tlsf_free(ptr2);
}
This work has been supported by the followin projects:
EUROPEAN: IST-2001-35102(OCERA) http://www.ocera.org.
SPANISH: TIN2005-08665-C3-03

3
lib/tlsf/SOURCES Normal file
View file

@ -0,0 +1,3 @@
#if CONFIG_CODEC == SWCODEC /* software codec platforms */
src/tlsf.c
#endif

9
lib/tlsf/TODO Normal file
View file

@ -0,0 +1,9 @@
To do list
==========
* Add mmap/sbrk support (DONE - V2.4).
* TLSF rounds-up request size to the head of a free list.
It has been shown to be a good policy for small blocks (<2048).
But for larger blocks this policy may cause excesive fragmentation.
A deeper analisys should be done.

34
lib/tlsf/libtlsf.make Normal file
View file

@ -0,0 +1,34 @@
# __________ __ ___.
# Open \______ \ ____ ____ | | _\_ |__ _______ ___
# Source | _// _ \_/ ___\| |/ /| __ \ / _ \ \/ /
# Jukebox | | ( <_> ) \___| < | \_\ ( <_> > < <
# Firmware |____|_ /\____/ \___ >__|_ \|___ /\____/__/\_ \
# \/ \/ \/ \/ \/
# $Id$
#
TLSFLIB_DIR := $(ROOTDIR)/lib/tlsf
TLSFLIB_SRC := $(call preprocess, $(TLSFLIB_DIR)/SOURCES)
TLSFLIB_OBJ := $(call c2obj, $(TLSFLIB_SRC))
TLSFLIB := $(BUILDDIR)/lib/libtlsf.a
OTHER_SRC += $(TLSFLIB_SRC)
INCLUDES += -I$(TLSFLIB_DIR)/src
EXTRA_LIBS += $(TLSFLIB)
TLSFLIBFLAGS = $(CFLAGS) -fstrict-aliasing -ffunction-sections $(SHARED_CFLAGS)
# Enable statistics in the sim
ifneq ($(findstring sdl-sim, $(APP_TYPE)), sdl-sim)
TLSFLIBFLAGS += -DTLSF_STATISTIC=1
endif
# special rules for tlsf
$(BUILDDIR)/lib/tlsf/src/%.o: $(TLSFLIB_DIR)/src/%.c
$(SILENT)mkdir -p $(dir $@)
$(call PRINTS,CC $(subst $(ROOTDIR)/,,$<))$(CC) -c $< -o $@ \
-I$(TLSFLIB_DIR)/src $(TLSFLIBFLAGS)
$(TLSFLIB): $(TLSFLIB_OBJ)
$(SILENT)$(shell rm -f $@)
$(call PRINTS,AR $(@F))$(AR) rcs $@ $^ >/dev/null

12
lib/tlsf/src/target.h Normal file
View file

@ -0,0 +1,12 @@
#ifndef _TARGET_H_
#define _TARGET_H_
#include <pthread.h>
#define TLSF_MLOCK_T pthread_mutex_t
#define TLSF_CREATE_LOCK(l) pthread_mutex_init (l, NULL)
#define TLSF_DESTROY_LOCK(l) pthread_mutex_destroy(l)
#define TLSF_ACQUIRE_LOCK(l) pthread_mutex_lock(l)
#define TLSF_RELEASE_LOCK(l) pthread_mutex_unlock(l)
#endif

1010
lib/tlsf/src/tlsf.c Normal file

File diff suppressed because it is too large Load diff

39
lib/tlsf/src/tlsf.h Normal file
View file

@ -0,0 +1,39 @@
/*
* Two Levels Segregate Fit memory allocator (TLSF)
* Version 2.4.4
*
* Written by Miguel Masmano Tello <mimastel@doctor.upv.es>
*
* Thanks to Ismael Ripoll for his suggestions and reviews
*
* Copyright (C) 2008, 2007, 2006, 2005, 2004
*
* This code is released using a dual license strategy: GPL/LGPL
* You can choose the licence that better fits your requirements.
*
* Released under the terms of the GNU General Public License Version 2.0
* Released under the terms of the GNU Lesser General Public License Version 2.1
*
*/
#ifndef _TLSF_H_
#define _TLSF_H_
#include <string.h> /* defines size_t */
extern size_t init_memory_pool(size_t, void *);
extern size_t get_used_size(void *);
extern size_t get_max_size(void *);
extern void destroy_memory_pool(void *);
extern size_t add_new_area(void *, size_t, void *);
extern void *malloc_ex(size_t, void *);
extern void free_ex(void *, void *);
extern void *realloc_ex(void *, size_t, void *);
extern void *calloc_ex(size_t, size_t, void *);
extern void *tlsf_malloc(size_t size);
extern void tlsf_free(void *ptr);
extern void *tlsf_realloc(void *ptr, size_t size);
extern void *tlsf_calloc(size_t nelem, size_t elem_size);
#endif