From: Denis Vlasenko <vda@ilport.com.ua>
To: Margit Schubert-While <margitsw@t-online.de>
Subject: Re: 2.6.13/14 x86 Makefile - Pentiums penalized ?
Date: Wed, 14 Sep 2005 10:54:08 +0300
User-Agent: KMail/1.8.2
Cc: chriswhite@gentoo.org, linux-kernel@vger.kernel.org,
       zwane@arm.linux.org.uk
References: <Pine.LNX.4.61.0509132345050.13185@montezuma.fsmlabs.com> <5.1.0.14.2.20050914092308.025ca630@pop.t-online.de>
In-Reply-To: <5.1.0.14.2.20050914092308.025ca630@pop.t-online.de>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Message-Id: <200509141054.08975.vda@ilport.com.ua>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1210
Lines: 29

> 
> In the Intel Architecture Optimization document it specifically states 
> (Chapter 2.6) :
> "Avoid instructions that unnecessarily introduce dependence-related
> stalls: inc and dec instructions, .....".
> And again on page 2-11 :
> "The inc and dec instructions should always be avoided. Using add and
> sub instructions instead avoids data dependence and improves performance".
> And on page 2-71 :
> "The inc and dec instructions modify only a subset of flags in the flag 
> register.
> This creates a dependence on all previous writes of the flag register.
> This is especially problematic when these instructions are on the critical
> path because they are used to change an address for a load on which
> many other instructions depend. "

Well it sounds dramatic but in reality I doubt it is such a big deal.

Anyone to bench it?

> However, the kernel include and arch have a liberal sprinkling of inc/dec,
> and AFAICT some of these in hot-path.
--
vda
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/