larry@utecfa.UUCP (Larry Philps) (04/23/85)
Here are our most recent set of benchmarks for the ns32016 now that we have
10MHz Rev N chips. The benchmark programs are those posted by Dan Ts'o of
Rockefeller University in September. They were run on the following systems:
ecfc vax780 8MBytes memory, FPA, Eagles on a Massbuss, 4.2BSD
hub vax750 6Mbytes memory, No FPA, Ra81's on uda50, 4.2BSD
sun SUN2/120 2Mbytes memory, the standard sun 140Mbyte disk, 4.2BSD
utopia ns32016 Rev N, 1Mbyte on-board memory, 10MHz, No wait states,
Eagle with Xylogics multibus controller, local 4.2BSD port
masscomp MC500 2Mbytes memory, Micropolis 50Meg winch on DSD5217,
enhanced System III UNIX.
The benchmark consisted of 10 tests which were:
CC-LOOP 1) Compile a 3 line C program.
LOOP 2) 1 million times around a tight loop.
CC-SIEVE 3) Compile a 30 line C program.
SIEVE 4) The sieve extraction algorithm that computed all the prime
numbers less than 8190 (done 10 times in each run).
GETPID 5) 100,000 getpid()'s.
NROFF 6) nroff -ms /dev/null.
GREP 7) grep zoom /usr/dict/words.
COPY 8) cp /usr/dict/words /tmp/words$$.
SORT 9) sort -r /usr/dict/words > /dev/null
FLOAT 10) 70,000 floating point multiplies and 70,000 floating point
divides.
Each test was run 3 times in succession on a single user system and the result
averaged. Each "time" is divided into 3 distinct numbers which represent: real,
user and system times respectively. There are 5 parts to the following table
which are:
1) A listing of the acquired data.
2) For each of the real, user, and system times a normalization to
that of the vax780 system.
(ie. (time for 780) / (time for system in question)
3) For each system the sum of the real, user, and system time for
ALL 10 tests was computed and then divided by the corresponding
total for the vax780. This results in the observed speed of
the system in terms of real, user, and system time relative
to a vax780.
4) The same as part 2 but relative to the vax750 system.
5) The same as part 3 but relative to the vax750 system.
Disclaimer: The results of these tests are provided for your information
only. I do not guarantee or even expect that you will be able to duplicate
them exactly. The numbers do not necessarily reflect my personal opinions
of the relative "goodness" of the various systems.
--------- cut here and print on 132 column line printer -------------
1) The actual times collected:
Test ecfc hub sun utopia masscomp
real user sys real user sys real user sys real user sys real user sys
---------+--------------------+-------------------+------------------+--------------------+-----------------
CC-LOOP 3.1 0.7 1.2 5.1 1.2 2.0 6.6 1.5 2.5 10.0 1.4 3.3 8.9 1.6 2.8
LOOP 2.5 2.5 0.0 5.0 4.9 0.0 6.8 6.6 0.1 8.0 7.9 0.1 6.5 6.2 0.2
CC-SIEVE 4.1 1.6 1.2 6.7 2.6 2.1 8.2 2.8 2.6 10.3 3.3 2.9 11.0 3.5 2.9
SIEVE 2.5 2.4 0.0 4.1 3.9 0.0 4.6 4.3 0.1 5.0 5.0 0.1 4.4 4.1 0.1
GETPID 15.8 2.5 13.2 24.8 4.8 19.9 26.2 3.5 22.6 29.3 6.7 22.8 25.3 1.2 23.8
NROFF 3.7 2.7 0.4 6.6 5.1 0.7 5.6 4.0 0.8 6.3 4.7 1.0 11.5 7.4 2.1
GREP 4.0 3.3 0.5 7.8 6.8 0.8 7.2 6.4 0.5 7.0 6.6 0.8 9.5 6.5 2.0
COPY 1.8 0.0 0.7 2.3 0.0 0.9 2.9 0.0 1.3 3.0 0.0 0.6 8.0 0.0 3.9
SORT 35.9 32.4 2.0 58.7 54.3 2.8 53.1 48.1 3.1 67.3 60.3 4.5 75.5 52.6 12.7
FLOAT 1.3 1.3 0.0 8.4 8.3 0.0 9.8 9.6 0.1 3.0 3.4 0.1 36.6 35.7 0.5
2) The times in part 1 normalized to that of the vax780 (ecfc):
Test ecfc hub sun utopia masscomp
real user sys real user sys real user sys real user sys real user sys
---------+--------------------+-------------------+------------------+--------------------+-----------------
CC-LOOP 1.00 1.00 1.00 0.61 0.58 0.60 0.47 0.47 0.48 0.31 0.50 0.36 0.35 0.44 0.43
LOOP 1.00 1.00 1.00 0.50 0.51 1.00 0.37 0.38 0.00 0.31 0.32 0.00 0.38 0.40 0.00
CC-SIEVE 1.00 1.00 1.00 0.61 0.62 0.57 0.50 0.57 0.46 0.40 0.48 0.41 0.37 0.46 0.41
SIEVE 1.00 1.00 1.00 0.61 0.62 1.00 0.54 0.56 0.00 0.50 0.48 0.00 0.57 0.59 0.00
GETPID 1.00 1.00 1.00 0.64 0.52 0.66 0.60 0.71 0.58 0.54 0.37 0.58 0.62 2.08 0.55
NROFF 1.00 1.00 1.00 0.56 0.53 0.57 0.66 0.68 0.50 0.59 0.57 0.40 0.32 0.36 0.19
GREP 1.00 1.00 1.00 0.51 0.49 0.62 0.56 0.52 1.00 0.57 0.50 0.62 0.42 0.51 0.25
COPY 1.00 1.00 1.00 0.78 1.00 0.78 0.62 1.00 0.54 0.60 1.00 1.17 0.22 1.00 0.18
SORT 1.00 1.00 1.00 0.61 0.60 0.71 0.68 0.67 0.65 0.53 0.54 0.44 0.48 0.62 0.16
FLOAT 1.00 1.00 1.00 0.15 0.16 1.00 0.13 0.14 0.00 0.43 0.38 0.00 0.04 0.04 0.00
3) The sum of each column in part 1 divided by the total of the corresponding column for the vax780
ecfc hub sun utopia masscomp
real user sys real user sys real user sys real user sys real user sys
Averages 1.00 1.00 1.00 0.58 0.54 0.66 0.57 0.57 0.57 0.50 0.50 0.53 0.38 0.42 0.38
4) The times in part 1 normalized to that of the vax750 (hub):
Test hub sun utopia masscomp
real user sys real user sys real user sys real user sys
---------+--------------------+-------------------+------------------+------------------
CC-LOOP 1.00 1.00 1.00 0.77 0.80 0.80 0.51 0.86 0.61 0.57 0.75 0.71
LOOP 1.00 1.00 1.00 0.74 0.74 0.00 0.63 0.62 0.00 0.77 0.79 0.00
CC-SIEVE 1.00 1.00 1.00 0.82 0.93 0.81 0.65 0.79 0.72 0.61 0.74 0.72
SIEVE 1.00 1.00 1.00 0.89 0.91 0.00 0.82 0.78 0.00 0.93 0.95 0.00
GETPID 1.00 1.00 1.00 0.95 1.37 0.88 0.85 0.72 0.87 0.98 4.00 0.84
NROFF 1.00 1.00 1.00 1.18 1.27 0.87 1.05 1.09 0.70 0.57 0.69 0.33
GREP 1.00 1.00 1.00 1.08 1.06 1.60 1.11 1.03 1.00 0.82 1.05 0.40
COPY 1.00 1.00 1.00 0.79 1.00 0.69 0.77 1.00 1.50 0.29 1.00 0.23
SORT 1.00 1.00 1.00 1.11 1.13 0.90 0.87 0.90 0.62 0.78 1.03 0.22
FLOAT 1.00 1.00 1.00 0.86 0.86 0.00 2.80 2.44 0.00 0.23 0.23 0.00
3) The sum of each column in part 1 divided by the total of the corresponding column for the vax750
hub sun utopia masscomp
real user sys real user sys real user sys real user sys
Averages 1.00 1.00 1.00 0.99 1.06 0.87 0.87 0.93 0.81 0.66 0.77 0.57
--
Larry Philps
Engineering Computing Facility
University of Toronto
{linus,ihnp4,uw-beaver,floyd,decvax,utzoo}!utcsri!utecfa!larry