From owner-freebsd-current  Fri Jan 26 07:41:55 1996
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id HAA09204
          for current-outgoing; Fri, 26 Jan 1996 07:41:55 -0800 (PST)
Received: (from dyson@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id HAA09194
          for current; Fri, 26 Jan 1996 07:41:52 -0800 (PST)
Date: Fri, 26 Jan 1996 07:41:52 -0800 (PST)
From: John Dyson <dyson>
Message-Id: <199601261541.HAA09194@freefall.freebsd.org>
To: current
Subject: Good news -- pipe stuff
Sender: owner-current@FreeBSD.ORG
Precedence: bulk


I am in the process of creating a new pipe mechanism.  Our pipes
have been a bit slower than they should be, considering the simple
functionality that they perform.  The latest results that I get
show significant performance improvements in both the Byte and lmbench
benchmarks on pipe latency, pipe bandwidth, and pipe context switch.

The stuff on my workstation includes this new code, and is isolated well
enough that it is "theoretically" possible to include even into -stable
(not likely though.)

Pipe latency is about 2x the speed, pipe bandwidth is about 50% faster,
and I am still evaluating the context switch -- but it is faster.  Before
I include the changes, I plan to add more intelligent memory management
of pipes (including making them pageable -- instead of tieing up wired
memory), and page-flipping.  Page-flipping should add about 50% more onto
the bandwidth.

Anyway, the code is simpler than I had first thought that it would be, and
should probably be ready for testing in the next week or so...

John
dyson@freebsd.org