From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([2001:4830:134:3::10]:37611)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1atf5B-00087L-TW
	for qemu-devel@nongnu.org; Fri, 22 Apr 2016 13:41:34 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1atf58-00061W-1d
	for qemu-devel@nongnu.org; Fri, 22 Apr 2016 13:41:33 -0400
Received: from mail-qk0-x243.google.com ([2607:f8b0:400d:c09::243]:36378)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <rth7680@gmail.com>) id 1atf57-00061S-T3
	for qemu-devel@nongnu.org; Fri, 22 Apr 2016 13:41:29 -0400
Received: by mail-qk0-x243.google.com with SMTP id l68so6252658qkf.3
	for <qemu-devel@nongnu.org>; Fri, 22 Apr 2016 10:41:29 -0700 (PDT)
Sender: Richard Henderson <rth7680@gmail.com>
References: <1461107270-19234-1-git-send-email-cota@braap.org>
	<1461107270-19234-12-git-send-email-cota@braap.org>
From: Richard Henderson <rth@twiddle.net>
Message-ID: <571A6245.4070209@twiddle.net>
Date: Fri, 22 Apr 2016 10:41:25 -0700
MIME-Version: 1.0
In-Reply-To: <1461107270-19234-12-git-send-email-cota@braap.org>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] [PATCH v3 11/11] translate-all: add tb hash bucket
 info to 'info jit' dump
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel/>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: "Emilio G. Cota" <cota@braap.org>, QEMU Developers <qemu-devel@nongnu.org>, MTTCG Devel <mttcg@greensocs.com>
Cc: =?UTF-8?Q?Alex_Benn=c3=a9e?= <alex.bennee@linaro.org>, Paolo Bonzini <pbonzini@redhat.com>, Peter Crosthwaite <crosthwaite.peter@gmail.com>, Peter Maydell <peter.maydell@linaro.org>, Sergey Fedorov <serge.fdrv@gmail.com>

On 04/19/2016 04:07 PM, Emilio G. Cota wrote:
> +    ht_avg_len = qht_avg_bucket_chain_length(&tcg_ctx.tb_ctx.htable, &ht_heads);
> +    cpu_fprintf(f, "TB hash avg chain   %0.5f buckets\n", ht_avg_len);
> +    cpu_fprintf(f, "TB hash size        %zu head buckets\n", ht_heads);


I think the accounting is questionable here.

Consider the following data:

TB count             230467/671088
TB invalidate count  25915
TB hash avg chain    1.03073 buckets
TB hash size         131072 head buckets

This means that we've got 230467 - 25915 = 204552 active TB's, installed into a
hash table with 131072 heads.  For a perfectly uniform distribution of TBs,
that would be an average chain length of 204552 / 131072 = 1.56.

In order to get the average down to 1.03, there need to be a substantial number
of heads with zero entries.

I think perhaps it might be more enlightening to separately account for empty
and non-empty heads.  E.g.

TB hash buckets used  xxxx/131072
TB hash avg chain     yyyy

where xxxx is the number of non-empty heads, and yyyy = |TBs| / xxxx.

I also wonder if it wouldn't be better to size the hash table as appropriate
for the maximum number of allowable TBs.


r~