Path: utzoo!utgpu!watmath!clyde!att!osu-cis!killer!gtmvax!dms3b1!dave
From: dave@dms3b1.UUCP (Dave Hanna)
Newsgroups: comp.sys.amiga
Subject: Re: CMI accelerator problems
Summary: 68881 speedup is not that great with a 68010, either
Keywords: CMI 68020 68881
Message-ID: <159@dms3b1.UUCP>
Date: 10 Nov 88 19:29:09 GMT
References: <6052@killer.DALLAS.TX.US>
Reply-To: dave@dms3b1.UUCP (Dave Hanna)
Distribution: na
Organization: Daltech MicroSystems, Dallas
Lines: 24

In article <6052@killer.DALLAS.TX.US> aimania@killer.DALLAS.TX.US (Walter Rothe) writes:
[stuff about Accellerator board cross talk with expansion boards deleted]
>By the way, the floating point speedup achieved from any 68881 connected to
>a 68k will not be as great as that you can get by connecting a 68881
>directly to a 68010 or 68020. There will still be a significant speedup on
               ^^^^^
>double precision floating point calculations but only a small( 2 or 3 times )
>speedup over the single precision Motorola FFP format.
   Minor point, but connecting a 68881 to either a 68000 or a 68010 will
not give you the speed up that it will on a 68020.  Both the 68000 and
68010 interface to the device as an I/O peripheral, rather than through
a co-processor interface, and there is a fairly significant amount of
overhead in setting up the calculation, particularly for simple operations.
Both the 68020 and 68030 use a co-processor interface, where 68881 operations
are simply another instruction to the CPU.

>Walter Rothe at the UNIX(Tm) Connection, Dallas, Tx


-- 
Dave Hanna,  Daltech MicroSystems    |  "Do or do not -- There is no try"
P.O. Box 584, Bedford, TX 76095      |                        - Yoda
(214) 358-4534   (817) 540-1524      |
UUCP:  ...!killer!gtmvax!dave        |