Date: Mon, 9 Aug 1999 14:37:55 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: pavenis AT lanet DOT lv, "Mark E." <snowball3 AT bigfoot DOT com>
cc: djgpp-workers AT delorie DOT com
Subject: Re: Performance regression of 2.95 vs. 2.7, x86, loop-related (fwd)
In-Reply-To: <B0000097614@stargate.astr.lu.lv>
Message-ID: <Pine.SUN.3.91.990809143435.2359E-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
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


On Mon, 9 Aug 1999 pavenis AT lanet DOT lv wrote:

> I tested DJGPP ports of gcc-2.7.2.1, gcc-2.8.1, egcs-1.1.2 and gcc-
> 2.95. The latest is ths only one which generates '.p2align 4,,7' (see 
> attachments).

Thanks.  I see that gcc 2.95 aligns function entry points on the 4-byte 
boundaries, and uses ".p2align 4,,7" for jump targets.

So it seems that the only thing we miss to have the better alignment is 
Binutils configured with 16-byte subsection alignment.

Mark, did you succeed to resolve the problems with doing this?