Path: utzoo!utgpu!watserv1!watmath!att!pacbell!pacbell.com!mips!zaphod.mps.ohio-state.edu!samsung!uunet!mcsun!ukc!inmos!roger@wraxall.inmos.co.uk
From: roger@wraxall.inmos.co.uk (Roger Shepherd)
Newsgroups: comp.arch
Subject: 64 by 32 divide needs 64 bit datapath?
Message-ID: <7254@ganymede.inmos.co.uk>
Date: 4 Jun 90 10:23:58 GMT
References: <9462@pt.cs.cmu.edu> <39213@mips.mips.COM>
Sender: news@inmos.co.uk
Reply-To: roger@inmos.co.uk (Roger Shepherd)
Organization: INMOS Limited, Bristol, UK.
Lines: 23

In article <39213@mips.mips.COM> mash@mips.COM (John Mashey) writes:

> However, in any case, it turns out that 64/32->32 requires a 64-bit
> wide datapath, which would have caused serious chip layout problems
> in the middle of something where every other integer thing was 32-wide.
> (32/32->32 is OK, and so is 32*32->64, i.e., / is a worse problem
> than *).

I'm sorry but I don't believe that a 64/32->32 requires a 64-bit wide 
datapath; in fact, I KNOW this isn't true since all our processors
implement double word/single word -> single word division on a 32-bit
datapath. 

So what gives?  I guess it must be either something to do with exactly
what is meant by a ``32-bit wide datapath'', or the fact that the rules
of VLSI design are different over here in Europe(:-). 



Roger Shepherd, INMOS Ltd   JANET:    roger@uk.co.inmos 
1000 Aztec West             UUCP:     ukc!inmos!roger or uunet!inmos-c!roger
Almondsbury                 INTERNET: roger@inmos.com
+44 454 616616              ROW:      roger@inmos.com OR roger@inmos.co.uk