From owner-freebsd-hardware  Tue Feb 25 01:21:59 1997
Return-Path: <owner-hardware>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id BAA12512
          for hardware-outgoing; Tue, 25 Feb 1997 01:21:59 -0800 (PST)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id BAA12507
          for <freebsd-hardware@freebsd.org>; Tue, 25 Feb 1997 01:21:56 -0800 (PST)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.8.3/8.6.9) id UAA03535; Tue, 25 Feb 1997 20:18:42 +1100
Date: Tue, 25 Feb 1997 20:18:42 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199702250918.UAA03535@godzilla.zeta.org.au>
To: bde@zeta.org.au, msmith@atrad.adelaide.edu.au
Subject: Re: Memory speed of P6-200 (256k)
Cc: freebsd-hardware@freebsd.org, ken@stox.pr.mcs.net
Sender: owner-hardware@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

>> Er, no, fast copying reduces buffer cache overheads.  Buffer cache
>> overheads are almost large enough to significantly reduce the advantages
>> of busmastering DMA over (slow) PIO.
>
>ie. the fast copyin/out code will make busmaster devices faster too? 8)

Yes, that was the original motivation for speeding up copyout() and
copyin().  Satoshi had lots of disk bandwidth (40MB/sec?) from multiple
controllers but couldn't use it all because copying alone was limited
to 40MB/sec.  He speeded it up to 70+MB/sec on a P5-Triton system by
using the FPU, and I speeded it up a few more MB/sec by fine tuning.
The dd speeds are almost twice as large since the input buffer is small
enough to stay in the P5's L1 data cache.

Bruce