Date: Tue, 2 Feb 1999 15:09:09 -0500
Message-Id: <199902022009.PAA04565@envy.delorie.com>
From: DJ Delorie <dj AT delorie DOT com>
To: djgpp AT delorie DOT com
In-reply-to: <3.0.6.32.19990202150512.0090dc30@pop.netaddress.com> (message
	from Paul Derbyshire on Tue, 02 Feb 1999 15:05:12 -0500)
Subject: Re: 64-bit integer math
References: <3 DOT 0 DOT 6 DOT 32 DOT 19990202125723 DOT 00904370 AT pop DOT netaddress DOT com>
 <Pine DOT GSO DOT 4 DOT 02 DOT 9902012103100 DOT 16025-100000 AT neptune DOT calstatela DOT edu>
 <3 DOT 0 DOT 6 DOT 32 DOT 19990202125723 DOT 00904370 AT pop DOT netaddress DOT com> <3 DOT 0 DOT 6 DOT 32 DOT 19990202150512 DOT 0090dc30 AT pop DOT netaddress DOT com>
Reply-To: djgpp AT delorie DOT com


> >The 387 can do 64-bit integer math, as all 64-bit integers can be
> >perfectly represented in 80-bit FP bit patterns.
> 
> I meant as integer operations. :-)

OK, I was close.  But, the 387 *can* be used as a 64-bit integer math
unit, as it has 64-bit int load/store, and if all the numbers and
results are 64-bit integers (i.e. be careful with division,
multiplication, arctangents, etc), you can treat it like a 64-bit
integer unit.

It's not as fast as simulating it in the CPU though.