Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!cmcl2!beta!hc!ames!amdcad!neptune!brian
From: brian@neptune.AMD.COM (Brian McMinn)
Newsgroups: comp.arch
Subject: Re: Sort Co-Processors
Message-ID: <426@neptune.AMD.COM>
Date: Fri, 9-Oct-87 15:49:41 EDT
Article-I.D.: neptune.426
Posted: Fri Oct  9 15:49:41 1987
Date-Received: Sun, 11-Oct-87 16:40:39 EDT
References: <112@sdeggo.UUCP>
Reply-To: brian@neptune.AMD.COM (Brian McMinn)
Organization: Advanced Micro Devices, Inc., Austin, Texas
Lines: 28

In article <112@sdeggo.UUCP> dave@sdeggo.UUCP (David L. Smith) writes:
>Has anyone out there ever run across a sorting co-processor?

This is not an idle idea, and there have been several sort coprocessors
proposed and a few built.  A few years ago, I was involved in the design
of the Am95C85 Content Addressable Data Manager (CADM for short).  The
CADM "is a microprocessor peripheral device capable of both storing
and managing data, thus relieving the host CPU of many time-consuming
data manipulation and management tasks."  (from the Tech Manual)

Each CADM contains 1K byte of data RAM and a micro control engine.
The CPU interface resembles a floating point co-processor: data and
commands are written to the CADM as an I/O device (or a special
address in memory, or ...).  Since 1K is not sufficient for most data
base applications, the micro control engine knows how to extend sort
and find operations across multiple cascaded CADMs.  This provides
the potential for up to 256 CADMs to operate as a single unit (but
capacitance considerations make 16 CADMs a much better choice for
speed).

For general company hype about the Am95C85, call your local field
sales office.  If that doesn't work well enough, call AMD Austin at
(512)-462-5334 and ask for Reed Borie.  He's our marketing wizard in
charge of CADM things.
-- 
	Brian McMinn		1-(512)-462-5389
	Advanced Micro Devices	domainLand:  brian@neptune.AMD.COM
	Austin, Texas		bangLand:   ...!amdcad!neptune!brian