Path: utzoo!utgpu!water!watmath!ccplumb
From: ccplumb@watmath.waterloo.edu (Colin Plumb)
Newsgroups: comp.arch
Subject: Re: Simulator vs Emulator
Message-ID: <16180@watmath.waterloo.edu>
Date: 10 Jan 88 01:34:03 GMT
References: <602@bnr-rsc.UUCP>
Reply-To: ccplumb@watmath.waterloo.edu (Colin Plumb)
Organization: U of Waterloo, Ontario
Lines: 30
Confusion: U. of Waterloo, Ontario

In article <602@bnr-rsc.UUCP> jim@bnr-rsc.UUCP (Jim Somerville) writes:
>I would simply like to know when to use the term "simulator" or
>"simulation" instead of "emulator" or "emulation".  Should the correct
>usage be, for example, "architecture simulator" or "architecture emulator"?
>"Instruction set simulator" or "instruction set emulator"?  Can these terms
>be completely interchanged?

Usually, if you don't have some feature, you can emulate it in
software.  The software does the job of the hardware.  A simulator is
used if you want to measure that something instead of using it.

Thus, you can't emulate H-bomb explosions, but you can simulate them.
Useful because it's cheaper and you can directly measure some events
which must be inferred from measurements taken of a real explosion.
You emulate a (co)processor if you want its functionality.  There are
programs to emulate IBM PC's and C-64's out there.  You simulate it if
you want to measure things about it (TLB miss rate, cycles/instruction,
etc.) like the processor architects do.

Another point is that emulators are (approximately) real-time, while
simulations are usually much slower.  (Weather prediction is an exception.)

Thus, if you care about the architecture, you probably want a
simulator.  If you only care about the instruction set, you probably
want an emulator.
--
	-Colin (watmath!ccplumb)

Zippy says:
Where do your SOCKS go when you lose them in th' WASHER?