From mboxrd@z Thu Jan 1 00:00:00 1970
Return-Path: <bruce.richardson@intel.com>
Received: from mga11.intel.com (mga11.intel.com [192.55.52.93])
by dpdk.org (Postfix) with ESMTP id C12BF7E23
for <dev@dpdk.org>; Wed, 10 Dec 2014 11:41:14 +0100 (CET)
Received: from fmsmga001.fm.intel.com ([10.253.24.23])
by fmsmga102.fm.intel.com with ESMTP; 10 Dec 2014 02:41:13 -0800
X-ExtLoop1: 1
X-IronPort-AV: E=Sophos;i="5.07,551,1413270000"; d="scan'208";a="635479097"
Received: from bricha3-mobl3.ger.corp.intel.com ([10.243.20.31])
by fmsmga001.fm.intel.com with SMTP; 10 Dec 2014 02:41:11 -0800
Received: by (sSMTP sendmail emulation); Wed, 10 Dec 2014 10:41:10 +0025
Date: Wed, 10 Dec 2014 10:41:10 +0000
From: Bruce Richardson <bruce.richardson@intel.com>
To: Michael Qiu <michael.qiu@intel.com>
Message-ID: <20141210104110.GB10056@bricha3-MOBL3>
References: <1418178341-4193-1-git-send-email-michael.qiu@intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1418178341-4193-1-git-send-email-michael.qiu@intel.com>
Organization: Intel Shannon Ltd.
User-Agent: Mutt/1.5.23 (2014-03-12)
Cc: dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH] Avoid possible memory cpoy when sort
hugepages
X-BeenThere: dev@dpdk.org
X-Mailman-Version: 2.1.15
Precedence: list
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>
X-List-Received-Date: Wed, 10 Dec 2014 10:41:15 -0000
On Wed, Dec 10, 2014 at 10:25:41AM +0800, Michael Qiu wrote:
> When the first address is the compared address in the loop,
> it will also do memory copy, which is meaningless,
> worse more, when hugepg_tbl is mostly in order. This should
> be a big deal in large hugepage memory systerm(like hunderd
> or thousand GB).
I actually doubt the time taken by this sorting is a significant part of the
initialization time taken, even for hundreds of GBs of memory. Do you have
any measurements to confirm this?
[However, that's not to say that we can't merge in this patch]
>
> Meanwhile smallest_idx never be a value of -1,so remove this check.
>
> This patch also includes some coding style fix.
>
> Signed-off-by: Michael Qiu <michael.qiu@intel.com>
> ---
> lib/librte_eal/linuxapp/eal/eal_memory.c | 13 +++++--------
> 1 file changed, 5 insertions(+), 8 deletions(-)
>
> diff --git a/lib/librte_eal/linuxapp/eal/eal_memory.c b/lib/librte_eal/linuxapp/eal/eal_memory.c
> index e6cb919..700aba2 100644
> --- a/lib/librte_eal/linuxapp/eal/eal_memory.c
> +++ b/lib/librte_eal/linuxapp/eal/eal_memory.c
> @@ -678,14 +678,13 @@ error:
> static int
> sort_by_physaddr(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
> {
> - unsigned i, j;
> - int compare_idx;
> + unsigned i, j, compare_idx;
> uint64_t compare_addr;
> struct hugepage_file tmp;
>
> for (i = 0; i < hpi->num_pages[0]; i++) {
> compare_addr = 0;
> - compare_idx = -1;
> + compare_idx = i;
>
> /*
> * browse all entries starting at 'i', and find the
> @@ -704,11 +703,9 @@ sort_by_physaddr(struct hugepage_file *hugepg_tbl, struct hugepage_info *hpi)
> }
> }
>
> - /* should not happen */
> - if (compare_idx == -1) {
> - RTE_LOG(ERR, EAL, "%s(): error in physaddr sorting\n", __func__);
> - return -1;
> - }
> + /* avoid memory copy when the first entry is the compared */
> + if (compare_idx == i)
> + continue;
>
> /* swap the 2 entries in the table */
> memcpy(&tmp, &hugepg_tbl[compare_idx],
> --
> 1.9.3
>