From: "Tim Van Holder" <tim DOT van DOT holder AT pandora DOT be>
To: <djgpp-workers AT delorie DOT com>
Subject: Re: memalign hooks (was: LIBM patch for GCC 3.3 - math changes)
Date: Wed, 25 Jun 2003 09:05:04 +0200
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit
X-Mailer: Microsoft Outlook, Build 11.0.4920
In-reply-to: <3EF8A24D.9BCD6C6F@phekda.freeserve.co.uk>
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.3718.0
Thread-Index: AcM6i5myj3mjrRoJTli+sGEefO7lPgAWvUaA
Message-Id: <20030625070545.B47E533DBA7@iceage.anubex.com>
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

I checked the glibc 2.3.2 code and it has this comment before the
memalign prototype:

/*
  memalign(size_t alignment, size_t n);
  Returns a pointer to a newly allocated chunk of n bytes, aligned
  in accord with the alignment argument.

  The alignment argument should be a power of two. If the argument is
  not a power of two, the nearest greater power is used.
  8-byte alignment is guaranteed by normal malloc calls, so don't
  bother calling memalign with an argument of 8 or less.

  Overreliance on memalign is a sure way to fragment space.
*/ 

The code agrees with this - if the requested alignment is less than
that provided by malloc(), the call is forwarded to malloc().
If it is smaller than the minimum size, the minimum size is used
instead (this seems an implementation detail of the malloc package
used by glibc).
Then the alignment size is brought up to a power of two, and normal
memalign operation proceeds.
memalign also has its own hook function (__memalign_hook) that is
called before anything else is done.

posix_memalign checks alignment and returns EINVAL instead of
increasing alignment and proceeding.