Re: Emulators

In article by emanuel stiebler: Here's my no-numbers-just-gut-feelings of the various PDP-11 emulators. John Wilson's Ersatz is probably the fastest; it's written in assembly code, and so gains a fair bit that way. Second would be the Begemot emulator. They've unrolled the instruction decode loop heavily, and that helps a lot. Bob Supnik's emulator would be the slowest of the three. However, it's still not that slow, may 1/3 the speed of Ersatz. Apout can't be compared to the above 3 emulators, because it doesn't emulate peripherals nor supervisor mode. User-mode instructions run at about the same speed as Supnik's emulator, but system calls are done by native code. The sole benchmark I have is: FreeBSD identifies my desktop box as Pentium II/Xeon/Celeron (348.93-MHz 686-class CPU). Using Apout, I can compile the 2.11BSD GENERIC kernel in 4 minutes 15 seconds. I'll try building Supnik and Begemot and getting comparative results. Last comment: all the simulators have strengths & shortcomings, and that applies not just to ease of use but also to CPU, I/O performance etc. You really have to try them all & find the one that suits you. Warren Received: (from major@localhost) by minnie.cs.adfa.edu.au (8.9.3/8.9.3) id NAA35744 for pups-liszt; Fri, 21 Jan 2000 13:44:42 +1100 (EST) Received: from begemot.org (negara.cs.waikato.ac.nz [130.217.248.112]) by minnie.cs.adfa.edu.au (8.9.3/8.9.3) with ESMTP id NAA35740 for &lt;pups(a)minnie.cs.adfa.edu.au&gt;au>; Fri, 21 Jan 2000 13:44:20 +1100 (EST) Received: (from joerg@localhost) by begemot.org (8.9.3/8.9.2) id PAA71802; Fri, 21 Jan 2000 15:45:24 +1300 (NZDT) (envelope-from joerg) Date: Fri, 21 Jan 2000 15:45:24 +1300 From: "Joerg B. Micheel" &lt;joerg(a)begemot.org&gt; To: Warren Toomey &lt;wkt(a)cs.adfa.edu.au&gt; Cc: Unix Heritage Society &lt;pups(a)minnie.cs.adfa.edu.au&gt;au>, joerg(a)begemot.org, "Hartmut B. Brandt" &lt;brandt(a)fokus.gmd.de&gt; Subject: Re: Emulators Message-ID: &lt;20000121154524.A71774(a)begemot.org&gt; References: <034201bf63ad$8ad414b0$5d01a8c0@p2350> &lt;200001210216.NAA53527(a)henry.cs.adfa.edu.au&gt; Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: &lt;200001210216.NAA53527(a)henry.cs.adfa.edu.au&gt;au>; from Warren Toomey on Fri, Jan 21, 2000 at 01:16:40PM +1100 Organization: Begemot Computer Associates Operating-System: ... powered by FreeBSD Sender: owner-pups(a)minnie.cs.adfa.edu.au Precedence: bulk On Fri, Jan 21, 2000 at 01:16:40PM +1100, Warren Toomey wrote: Us too! :-) The problem is that it doesn't scale that simple. Each and every instruction has the parsing overhead. Next comes execution overhead. You'll find that the parsing is pretty constant, no matter whether it is a NOP or some sophisticated MUL command. The execution speed varies heavily, very often it is alot faster than the original hardware. IO has seen a tremendous speedup, we can benefit here from todays hardware alot. Just remeber how long it took to get a prompt or echo when hitting the keyboard. As a result, the original feeling of the real machine is lost, very unfortunate. As a rough summary, simple commands do not improve (much), whereas everything complex speeds up with the emulator. Harti has done quite a bit of testing on different instructions and compared them to an LSI11/73 (KDJ11A). The emulator was run on a i486 at the time. Have a look at the p11 distribution, it should be in Tests somewhere. Regards, Joerg -- Joerg B. Micheel Email: &lt;joerg(a)begemot.org&gt; Begemot Computer Associates Phone: +64 7 8562148 40 Masters Avenue, Hillcrest Fax: +64 7 8562148 Hamilton, New Zealand Pager: +64 868 38222 Received: (from major@localhost) by minnie.cs.adfa.edu.au (8.9.3/8.9.3) id DAA39169 for pups-liszt; Sat, 22 Jan 2000 03:08:07 +1100 (EST)