From:	"A.Appleyard" <A DOT APPLEYARD AT fs2 DOT mt DOT umist DOT ac DOT uk>
Organization:	Materials Science Centre
To:	DJGPP AT DELORIE DOT COM
Date:	Fri, 14 Mar 1997 15:53:13 GMT
Subject:	A long-standing oddity in malloc()
Message-ID:	<4D54BF3B83@fs2.mt.umist.ac.uk>

  (Please also reply personally to a DOT appleyard AT fs2 DOT mt DOT umist DOT ac DOT uk)

  In the file djgpp\src\libc\ansi\stdlib\malloc.c, on working manually through
the source forms of malloc() and its helper morecore():-
  (0) Here, OV = sizeof(union overhead).
  (1) The first call of malloc() has 11 lines of text which carefully align
the sbrk() at OV bytes less than a page boundary. This is clearly so that as
few page boundaries as possible occur inside allocated store blocks; but what
harm befalls if allocated blocks straddle page boundaries?
  (2) If I ask for a block of size N, if N+OV > pagesize, having rounded N up
to P (where P = the next (OV bytes less than a multiple of pagesize)),
morecore() asks for not (P/pagesize) pages but (P/pagesize+1) pages!!! Why the
unused extra page each time??? If many of the malloc() requests are from 2 to
a few pages each, a lot of memory is going to be wasted in these unused pages.

- Raw text -