Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!sdd.hp.com!elroy.jpl.nasa.gov!ncar!gatech!prism!mailer.cc.fsu.edu!uflorida!travis!hardy!dana
From: dana@hardy.hdw.csd.harris.com (Dan Aksel)
Newsgroups: comp.arch
Subject: Re: IEEE arithmetic (Goldberg paper)
Message-ID: <3646@travis.csd.harris.com>
Date: 12 Jun 91 13:12:30 GMT
References: <9106112357.AA18157@ucbvax.Berkeley.EDU>
Sender: news@travis.csd.harris.com
Organization: Harris Computer Systems, Ft. Lauderdale, FL
Lines: 20

In article <9106112357.AA18157@ucbvax.Berkeley.EDU> jbs@WATSON.IBM.COM writes:
>
>          I said:
>>If your chip has 80 bit registers I fail to see how you can avoid
>>performing 80 bit loads and stores while saving and restoring these registers.
>-----------------------------------------------------------
>          Perhaps not.  Of course one way people have managed in the past
>is to only save and restore the first 64 bits of the register.  In my
>view this demonstrates excessive imagination.
>                          James B. Shearer
Solution is relatively simple.  Hardware provides a status register which
indicates the precision of each register.  A computed GOTO tells which format to
save each individual register in.  This may create too much overhead if a rapid
context switch time is the critical path instead of memory space.

THe only compromise I can see immediately is a two bit status field which tells
the precision of the LARGEST format currently stored in the register file.  All
registers are then saved in this format.  Not an optimal solution but it works.
---
Dan AKsel