From mboxrd@z Thu Jan  1 00:00:00 1970
From: Stephen Hemminger <stephen@networkplumber.org>
Subject: Re: [PATCH] Performance optimization of ACL build process
Date: Wed, 31 Aug 2016 08:24:56 -0700
Message-ID: <20160831082456.21423b42@xeon-e3>
References: <20160816140128.10149-1-vladyslav.buslov@harmonicinc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Cc: <konstantin.ananyev@intel.com>, <dev@dpdk.org>
To: Vladyslav Buslov <vladyslav.buslov@harmonicinc.com>
Return-path: <dev-bounces@dpdk.org>
Received: from mail-pa0-f43.google.com (mail-pa0-f43.google.com
 [209.85.220.43]) by dpdk.org (Postfix) with ESMTP id B0E652946
 for <dev@dpdk.org>; Wed, 31 Aug 2016 17:24:47 +0200 (CEST)
Received: by mail-pa0-f43.google.com with SMTP id cy9so19701002pac.0
 for <dev@dpdk.org>; Wed, 31 Aug 2016 08:24:47 -0700 (PDT)
In-Reply-To: <20160816140128.10149-1-vladyslav.buslov@harmonicinc.com>
List-Id: patches and discussions about DPDK <dev.dpdk.org>
List-Unsubscribe: <http://dpdk.org/ml/options/dev>,
 <mailto:dev-request@dpdk.org?subject=unsubscribe>
List-Archive: <http://dpdk.org/ml/archives/dev/>
List-Post: <mailto:dev@dpdk.org>
List-Help: <mailto:dev-request@dpdk.org?subject=help>
List-Subscribe: <http://dpdk.org/ml/listinfo/dev>,
 <mailto:dev-request@dpdk.org?subject=subscribe>
Errors-To: dev-bounces@dpdk.org
Sender: "dev" <dev-bounces@dpdk.org>

On Tue, 16 Aug 2016 17:01:27 +0300
Vladyslav Buslov <vladyslav.buslov@harmonicinc.com> wrote:

> Hello,
> 
> In our application we need to be able to allocate tens of thousands of ACLs at runtime.
> Testing revealed significant performance problems. We were able to track them to memset in calloc function which caused multiple page walks per invocation.
> Modifying tb_mem to use huge page memory resulted ~2x performance gain for that operation.
> 
> Regards,
> Vladyslav
> 
> Vladyslav Buslov (1):
>   acl: use rte_calloc for temporary memory allocation
> 
>  lib/librte_acl/tb_mem.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 

What about making sure that the ACL code does it's own initialization of all fields
and just use malloc rather than calloc?