Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: notesfiles
Path: utzoo!watmath!clyde!cbosgd!ihnp4!qantel!hplabs!hp-pcd!hpfcla!tim
From: tim@hpfcla.UUCP (tim)
Newsgroups: net.micro
Subject: Re: Orphaned Response
Message-ID: <10100005@hpfclp.UUCP>
Date: Wed, 31-Jul-85 17:34:00 EDT
Article-I.D.: hpfclp.10100005
Posted: Wed Jul 31 17:34:00 1985
Date-Received: Sun, 4-Aug-85 10:41:29 EDT
References: <142@brl-tgr.UUCP>
Organization: Hewlett-Packard - Fort Collins, CO
Lines: 25
Nf-ID: #R:brl-tgr:142:hpfclp:10100005:37777777600:1353
Nf-From: hpfclp!tim    Jul 31 13:34:00 1985


Re: 9845B math co-processor

The architecture of the 9845B is a dual processor system.  One processor is
the 'OS' and 'I/O' engine and the other processor is a 'language' processor.
The math is performed (in most situations) in the 'language' processor.  These
processors are homogeneous 16-bit micro-processors.  There is an upgrade 
option for the 'language' processor to convert it from the 'micro' to a 
bit-sliced emulation of the processor.  You should be able to check with 
your local HP sales office about this option.  The math in the 'micro' is
done in BCD (the 'micro' had 3 pieces - an I/O sub processor, a regualar
processor, and a math sub processor).

The dual-CPU architecture was pretty interesting and you could do some neat
things with it.  The dual-CPU aspect was one of the reasons you could have
'SERIAL' and 'OVERLAP' statements in BASIC.  Normal mode was serial - where
the language processor would wait for any I/O a BASIC line did to finish before
going onto the next line of BASIC.  When you put it in overlap mode, the 
language processor would go on to the next line.  The underlying system had
'busy bits' that allowed the system to keep track of which variables had
been used in an I/O operation.  This would cause the language processor to 
wait before mucking with a busy variable.  

Tim Mikkelsen
hplabs!hpfcla!tim