Content-Type: text/plain; charset=US-ASCII
From: Andrew Theurer <habanero@us.ibm.com>
Reply-To: habanero@us.ibm.com
To: linux-kernel@vger.kernel.org
Subject: Re: must-fix list, v5
Date: Wed, 21 May 2003 20:25:47 -0500
User-Agent: KMail/1.4.3
References: <20030521152255.4aa32fba.akpm@digeo.com>
In-Reply-To: <20030521152255.4aa32fba.akpm@digeo.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 7BIT
Message-Id: <200305212025.47875.habanero@us.ibm.com>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1102
Lines: 25

On Wednesday 21 May 2003 17:22, Andrew Morton wrote:
> +o Overeager affinity in presence of repeated yields
> +
> +  http://www.hpl.hp.com/research/linux/kernel/o1-openmp.php
> +
> +  ingo: this is valid.  fix is in progress.

I have seen this with SpecJBB on high warehouse runs in around 2.5.66, I am 
testing now to see if it still exists in 2.5.69.  In 2.5.66 adding a bit to 
try_to_wakeup to relocate p-task_cpu to an idle cpu (if available) if the 
p->task_cpu is currently non idle helped dramatically.  This change is also 
in Andrea's 2.4 kernel tree I believe.  Basically in some situations, a quick 
push load balance before task activation.  

Also, in this particular case, this behavior may go away with the use of 
futexes in JVM, not sure yet, so I don't know if JVM and JBB are good 
justifications to make this change.  

-Andrew Theurer

-
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/