Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxl!eagle!harpo!ihnp4!inuxc!pur-ee!uiucdcs!uiucuxc!beaucham
From: beaucham@uiucuxc.UUCP
Newsgroups: net.micro.16k
Subject: Re: LMC 16032: Any - (nf)
Message-ID: <6232@uiucdcs.UUCP>
Date: Sat, 17-Mar-84 22:48:17 EST
Article-I.D.: uiucdcs.6232
Posted: Sat Mar 17 22:48:17 1984
Date-Received: Mon, 19-Mar-84 06:56:18 EST
Lines: 37

#R:hao:-86400:uiucuxc:25800001:000:2645
uiucuxc!beaucham    Mar 17 14:07:00 1984

I visited LMC in Chicago yesterday and was very impressed with their
machine, but had some reservations.  We brought with us an 8" floppy
created by tar on our vax780 holding three short source programs, two
f77 and one c; the floppy was in IBM standard SS/SD format.  They copied
the 8" to a 5.25", inserted this into the MegaMicro, and promptly installed
the three files. (I don't know what they used to make the copy.)  All three
programs compiled without a hitch.  At first, one of the two ftn programs
would not execute, and we thought at first it was due to incorrect compilation
of a COMPLEX function; however, it turned out to be a bad FPU chip, which when
replaced resulted in perfect execution of the program.
   I was interested in benchmarking ftn and c in comparison to our vax780.
The results using %time ~cbin~ were 13.1u + 0.5s = 13.6 vax780
                                    56.7u + 3.1s = 59.8 LMC
and after LMC engineer Robert Farber performed some optimization on the
c code and the resulting assembly (in anticipation of an improved c compiler)
the later result improved to        42.0u + 2.7s = 44.7 LMC
So on c execution the ratio seems to be about 3 or 4 to 1.
   With Fortran things are not so impressive.  For one program the comparison
was        0.3u + 0.1s = 0.4 vax780; 2.0u + 6.5s = 9.0 LMC >> 20 to 1
andtheother .5u + 0.3s = 0.8 vax780; 2.5u + 5.7s = 8.2 LMC >> 10 to 1.
Unfortunately, we did not time the ftn & c compilations, but since we were
the only ones on the LMC i would estimate a 20 to 1 ratio to the vax.  In
summary it seems that they are capable of 4 to 1 ratios to the vax with c,
but for ftn are they any faster than an IBM pc with an 8087?
     The package looks very nice and compact.  The unit they demoed had the
dumb serial interface but they strongly recommend the smart one.  The box
will hold 15 multibus cards;i wonder if the fan could handle 15 cards.However,

  with about six cards in it, it was running very cool and surprisingly quiet (about 60 DBA).  One fear i have concerns the sensitivity of the National chips.Evidentally, the supply must be adjusted with +-.1 volt for proper operation andthe chips must be selected to have identical best running voltages.  During our demo we had one crash and an engineer had to adjust the voltage for best operation.  They claimed this particular unit was hastily put together because their good units were out for field demos.  
      i think the LMC will handle 4 users without problem.  I too am looking
for competitive machines. 

             James Beauchamp, music bldg., uiuc, urbana, IL 61801
             (217) 344-3307/333-1089