Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site scirtp.UUCP
Path: utzoo!watmath!clyde!cbosgd!ihnp4!mhuxn!mhuxr!mhuxt!houxm!whuxl!whuxlm!akgua!mcnc!rti-sel!scirtp!dfh
From: dfh@scirtp.UUCP (David F. Hinnant)
Newsgroups: net.arch
Subject: how to outrun the prefetch queue?
Message-ID: <559@scirtp.UUCP>
Date: Mon, 30-Dec-85 14:38:31 EST
Article-I.D.: scirtp.559
Posted: Mon Dec 30 14:38:31 1985
Date-Received: Wed, 1-Jan-86 00:54:14 EST
Distribution: net
Organization: SCI Systems, Research Triangle Park, NC
Lines: 28

For a particular benchmark test I'm writing, I need to outrun the
prefetch queue of various microprocessors.  With Intel architecture,
you can repeat something like:

	mov ax, ax
  	mov ax, ax
  	mov ax, ax
  	...

and soon outrun the prefetch queue easily on the 8086, 80186, and 80286
since it takes longer to load the instruction sequence than to execute it
(four ticks to load the instruction (best case) (three for the 80286) and
two ticks to execute it). What instruction(s) can I use with the Motorola,
National Semi, and Zilog families to outrun the prefetch queue?  From the 
data manuals I have on the 68000, it looks like a:

	move d1, d2
  	move d1, d2
 	move d1, d2
  	...

will keep up with the prefetch queue, but not outrun it.  Any comments
appreciated.

-- 
				David Hinnant
				SCI Systems, Inc.
				...{decvax, akgua}!mcnc!rti-sel!scirtp!dfh