From: utzoo!decvax!harpo!utah-cs!utah-gr!thomas
Newsgroups: net.unix-wizards
Title: Re: compaction using realloc
Article-I.D.: utah-gr.663
Posted: Mon Dec 13 10:23:15 1982
Received: Tue Dec 14 05:13:46 1982
References: sri-unix.4783

I have a version of malloc (which I got from somebody else) which we
are using on our Vaxen now.  It satisfies all memory requests by
allocating a block which is the next power of 2 bigger than the request.
This allows it to keep separate free lists for each size of request,
making malloc() much faster on a paging machine (it only needs to keep
29 free lists, most of which are never used).  Obviously it depends
on the fact that memory is cheap on a Vax.  We have a program which used
to go into horrendous thrashing as soon as its data space size reached
the amount of real memory on the system, do to the standard malloc's
behaviour of searching all blocks to find a free one.  Our malloc also
has a little instrumentation built into it, so you can easily find
"hardening of the arteries" if you are interested.  I have permission
from the author to distribute it, if anybody is interested, drop
me a line.

=Spencer (harpo!utah-cs!thomas, thomas@utah-20)