donn@utah-cs.UUCP (Donn Seeley) (06/18/86)
I received this message from Prof. Carter recently -- it covers his
experiments with the Spice circuit simulator and the new 'fort'
compiler from DEC. Looks pretty good. BTW, in case anybody asks, the
compiler itself was only run on licensed machines; I mention this
because the compiler and all related utilities print an obnoxious
warning message when run under 4.n BSD.
Donn Seeley University of Utah CS Dept donn@utah-cs.arpa
40 46' 6"N 111 50' 34"W (801) 581-5668 decvax!utah-cs!donn
------------------------------------------------------------------------
Date: Mon, 16 Jun 86 19:20:56 mdt
From: carter@silicon (Tony M. Carter)
To: donn@UTAH-CS
Subject: spice comparisons
Performance is measured relative to the Gould 9080 using -lnm.
The "average time" was computed by adding the CPU time for the Utah and
Berkeley circuits and dividing by 2. The surprising thing is that the
Gould and the 8600 are exactly the same speed on the average (for these
two problems)!
The new f77 compiler on the vaxes resulted in a speedup of ~14%, while the
-lnm library on the gould resulted in a speedup of ~12% (both for utah.ckt).
The old version of spice on the 8600 was discarded long ago, and so no
improvement measures on bkly.ckt can be calculated.
Another interesting item: the new f77 compiler apparently sped up the
simulation of bkly.ckt more than utah.ckt. Note that the vaxen are the only
machines which ran bkly.ckt significantly faster than utah.ckt!
VMS fortran sped up the 8600 version by 1.36, and the others by about 1.7.
Machine Name Spice Time Rel. Spice Time Rel. Avg. Time Rel.
(Utah) Perf. (Bkly) Perf. Perf.
---------------------------------------------------------------------------
VAX 8600 (utah-cs) VMS for 22.07 1.000 19.93 1.000 21.00 1.000
Gould (utah-vlsi) -lnm 30.03 0.735 30.53 0.653 30.28 0.694
VAX 8600 (utah-cs) new f77 32.73 0.674 27.83 0.716 30.28 0.694
VAX 785 (utah-gr) VMS for 48.37 0.456 43.22 0.461 45.80 0.459
VAX 780 (utah-ug) VMS for 74.88 0.295 65.53 0.304 70.21 0.299
DEC 20 (utah-20) 80.17 0.275 68.95 0.289 74.56 0.282
VAX 785 (utah-gr) new f77 82.25 0.268 71.77 0.278 77.01 0.273
VAX uII (utah-zion) VMS for 83.97 0.263 75.43 0.264 79.70 0.263
VAX 750 (utah-muddy) VMS for 109.87 0.201 96.35 0.207 103.11 0.204
DN 160 (utah-cycles 1f2c) 104.00 0.212 115.00 0.173 109.50 0.192
VAX 780 (utah-ug) new f77 121.82 0.181 107.87 0.185 114.85 0.183
VAX uII (utah-zion) new f77 134.62 0.164 119.63 0.167 127.13 0.165
DN 330 (utah-kloo) 170.00 0.130 169.00 0.118 169.50 0.124
VAX 750 (utah-orion) new f77 182.67 0.121 159.83 0.125 171.25 0.123
DN 320 (utah-kzin 1562) 265.00 0.083 263.00 0.076 264.00 0.080
HP 9300 (utah-carter) -l881 269.37 0.082 299.12 0.067 284.25 0.074
SUN (utah-yampa) -fsky 605.40 0.036 778.00 0.026 691.50 0.030
DN 300* (utah-cmos 0802) 865.00 0.026 1007.00 0.020 936.00 0.022
DN 550* (utah-blue 4a23) 959.00 0.023 1125.00 0.018 1042.00 0.020
IBM PC/XT [PSPICE] 1077.79 0.020 FAILED ***** ******* *****
NOTE: Benchmark files are:
utah-vlsi:/vlsi/misc/src/spice/utah.ckt [cmos schmitt trigger]
utah-vlsi:/vlsi/misc/src/spice/bkly.ckt [nmos amplifier]