Date: Tue, 8 Jun 1999 10:18:45 -0400
Message-Id: <199906081418.KAA21495@envy.delorie.com>
From: DJ Delorie <dj AT delorie DOT com>
To: djgpp-workers AT delorie DOT com
CC: pavenis AT lanet DOT lv, djgpp-workers AT delorie DOT com
In-reply-to: <Pine.SUN.3.91.990608112824.4197B-100000@is> (message from Eli
	Zaretskii on Tue, 8 Jun 1999 11:28:44 +0300 (IDT))
Subject: Re: egcs-19990602 (gcc-2.95 prerelease) binaries for testing
References:  <Pine DOT SUN DOT 3 DOT 91 DOT 990608112824 DOT 4197B-100000 AT is>
Reply-To: djgpp-workers AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk


> Some time ago we had a thread here about alignment of data sections in
> DJGPP executables.  As far as I remember, the conclusion was that we
> want to increase the alignment to 32 bytes.  Does this have any
> relation to the compiler (as opposed to Binutils), and if so, does
> this snapshot do the Right Thing?

Alignment requires help from the compiler and the linker, although I
suspect the linker will do the right thing by default (align the
sections themselves).

8 bytes seems to be where people are heading.  Going further doesn't
add much to performance, but wastes a lot of cycles due to the larger
size of the memory space.  We currently use 4 bytes.  Note that this
is for a section-level alignment; things like ints and pointers would
still only need 4-byte alignment, but we'd want long longs and doubles
to be 8-byte aligned.