From owner-freebsd-current  Tue Jul 13  3:39:15 1999
Delivered-To: freebsd-current@freebsd.org
Received: from alcanet.com.au (border.alcanet.com.au [203.62.196.10])
	by hub.freebsd.org (Postfix) with ESMTP id B365114D70
	for <freebsd-current@FreeBSD.ORG>; Tue, 13 Jul 1999 03:39:11 -0700 (PDT)
	(envelope-from jeremyp@gsmx07.alcatel.com.au)
Received: by border.alcanet.com.au id <40341>; Tue, 13 Jul 1999 20:20:11 +1000
Date: Tue, 13 Jul 1999 20:38:01 +1000
From: Peter Jeremy <jeremyp@gsmx07.alcatel.com.au>
Subject: LOCK overheads (was Re: "objtrm" problem probably found)
To: dillon@apollo.backplane.com, freebsd-current@FreeBSD.ORG
Message-Id: <99Jul13.202011est.40341@border.alcanet.com.au>
Sender: owner-freebsd-current@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.ORG

To expand on this morning's table:

        i386SX-25 486DX2-50  P-133   PII-266  PIII-450  nproc  locks
mode  0   1950.23   161.17    39.65    26.31     9.21     EMPTY
mode  1   3340.59   221.74    71.74    24.45    16.48     1      no  tight
mode  2   3237.57   222.96    71.18    25.27    23.65     2      no  tight
mode  3   3367.65   626.28   282.31   153.29    93.02     1     yes  tight
mode  4   3263.64   627.83   285.58   152.94   160.82     2     yes  tight
mode  5   9439.15  2168.12   446.16    60.40    37.64     1      no  spread
mode  6  10231.96  2170.48   467.39    60.16    89.28     2      no  spread
mode  7  10660.05  3271.40   725.80   153.18    88.32     1     yes  spread
mode  8   9990.18  2243.56   755.87   155.18   161.08     2     yes  spread

mode  9   5544.82   623.13   131.31    49.96        ?     EMPTY
mode 10   7234.97   705.60   174.20    64.81        ?     1      no  tight
mode 11   7212.14   707.13   178.72    64.87        ?     2      no  tight
mode 12   7355.46  1075.49   304.74   182.75        ?     1     yes  tight
mode 13   6956.54  1076.95   327.11   180.21        ?     2     yes  tight
mode 14  13603.72  2486.87   582.02   100.10        ?     1      no  spread
mode 15  13443.54  1808.75   543.97   101.13        ?     2      no  spread
mode 16  13731.94  1587.16   717.31   207.12        ?     1     yes  spread
mode 17  13379.62  2547.45   800.31   207.70        ?     2     yes  spread

This suggests that adding a lock prefix to anything more recent than
a 386 is fairly expensive.  Since the 486 book states `1 clock', but
a real example takes more like 20 clocks, it seems that external
factors outweigh the stated execution timings.

Peter


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message