caf@omen.UUCP (Chuck Forsberg WA7KGX) (02/13/86)
*** REPLACE THIS LINE WITH YOUR BENCHMARK *** A Quickie Benchmark for Unix Systems Revised Feb-12-86 Chuck Forsberg Omen Technology INC The following benchmark was recommended by some folks from National Semiconductor as they felt the siev benchmark with its tight loops "discriminated" againgst highly pipelined CPU's. It's also very easy to type in, and catches some of the botches in alleged "clones" of Unix(TM). On unbundled systems, bc and an editor are almost always included in the "bacic package", so lack of a C compiler need not preclude running this benchmark. If the system cannot run this benchmark, it probabaly lacks something else YOU need to get your work done. Unless otherwise noted, the file "bct" was created with vi, chmod executable, and run as the shall script shown. Although not stated here, you can get a good idead of a system's responsiveness by going through these steps. Bct gives a reasonable indication of how quickly a machine processes compute intensive tasks such as troff and compiler passes. DISCLAIMER: Vendors are constntly improving their products by making them both faster and more compatible with the Real Unix. If you see a machine with bad results, consider it a warning and not condemnation. time bc <<f 2 ^ 4096 f Make bct executable. Clear the screen (no scrolling please). Then run it. Real Time System/comments 0:06.6 Gould UTX32 6/84 0:18 2.8 BSD PDP 11/70 1983 0:29 Sun 3/52 15 mHz 1ws 68020 2/86 0:31 Intel 310 + UniPort Unix SYS V, 2/86 0:33 PC-AT 9.05 mHz + SCO SYS V Xenix 2/86 0:36 PC-AT,Sritek 68020 prototype many ws 2/86 0:41 Universe 32 2/86 0:42 Pyramid w/o cache 6/84 0:48 4.2 BSD VAX 11/780 7/84 0:51 Intel 310 + Xenix 6/84 1:18 PC-XT Coherent (ed, no vi) 1983 * 1:20 IBM PC-RT + Unix, etc. 2/86 1:21 Plexus P25 (5 mHz) SYS 3 1983 1:34 Sun 68010 4.2 BSD 6/84 1:44 UNOS Charles River (No vi; no << in shell) * 6/84 1:51 Callan Unistar 300 6/84 1:59 Gould Powerstation 2000 (about $7k) 6/84 2:39 Cyb Work Mgr II 6/84 prog uses slow memory 6/84 2:50 LMC 16032 6mHz/3ws Target: 10mHz no ws 6/84 3:04 PCIX 6/84 (No vi) 3:15 PC-XT Venix 6/84 3:35 PC-XT SCO XENIX 6/84 3:50 SCO XENIX 6/84 4:(Approximate) Unisoft SYS 5 Lisa II (time command lied) 6/84 --- REGULUS (Wrong Answer; no << in "shell"; strange ed, no vi) * 6/84 * Independent versions of bc affect the timings.
henry@utzoo.UUCP (Henry Spencer) (02/21/86)
> The following benchmark was recommended by some folks from National > Semiconductor as they felt the siev benchmark with its tight loops > "discriminated" againgst highly pipelined CPU's... What this means, of course, is that highly pipelined CPUs have trouble running almost anything whose speed is primarily determined by tight loops. There are a LOT of such programs. One can argue that the Sieve benchmark is pointing out a real problem, and that this "discrimination" is legitimate. -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry