rockbox

History

Daniel Stenberg 1dd21edacf malloc(), best-fit for big blocks, small blocks treated separately, all details in THOUGHTS. No headers and stuff added yet. git-svn-id: svn://svn.rockbox.org/rockbox/trunk@569 a1c6a512-1295-4272-9138-f99709370657		2002-05-13 19:35:10 +00:00
..
bmalloc.c	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
bmalloc.h	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
bysize.c	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
bysize.h	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
dmalloc.c	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
dmalloc.h	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
dmytest.c	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
Makefile	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
Malloc.c	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
mytest.c	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
README	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00
THOUGHTS	malloc(), best-fit for big blocks, small blocks treated separately, all	2002-05-13 19:35:10 +00:00

README

Package: dbestfit - a dynamic best-fit memory allocator
Date:	 1996 - 2002
Version: 3.3
Author:	 Daniel Stenberg <daniel@haxx.se>
License: MIT

 I wrote the dmalloc part for small allocation sizes to improve the behavior
of the built-in (first-fit) allocator found in pSOS, during late 1996 and
spring 1997.

 I wrote the bmalloc part (best-fit with optional splay-tree sorting) just for
the fun of it and to see how good malloc() implementation I could make. The
quality of my implementation is still left to be judged in real-world tests.

TODO:
 * Remove the final not-so-very-nice loop in dmalloc.c that checks for a block
   with free fragments (when the list gets longer too much time might be spent
   in that loop).

 * Make a separate application that samples the memory usage of a program
   and is capable of replaying it (in order to test properly).