Java Microbenchmark - Old Results

[ index | benchmarks | speed | size | maintainability | rules | tools | compilers ]

v1.01, August 4

The first released version of the benchmarks. JIT compilers could hide cheap operations such as integer assignment and increment in empty slots in the timing loop, resulting in meaninglessly low benchmark times for those operations (marked with a *).

Browser CPU
& speed		 Null
Loop		 Local
i=n		 this.
i=n		 Array
a[0]=n		 Byte
b++		 Short
s++		 Int
i++		 Long
l++		 Float
f++		 Doub.
d++		 New
obj.		 Call
f()		 Sync
f()		 Math
abs()		 Inline
abs()
NN
3.0b6		 80486
120 MHz		 0.059 0.007 0.043 0.110 0.057 0.057 0.025 0.049 0.180 0.190 30.0 0.180 17.0 0.52 0.120
IE
3.0b2		 80486
120 MHz		 0.066 0.00* 0.009 0.034 0.033 0.033 0.00* 0.040 0.160 0.190 14.0 0.130 3.8 0.46 0.040
NN
2.02		 Pentium
133 MHz		 1.100 0.430 0.720 1.100 1.200 1.200 0.250 1.100 1.200 1.000 21.0 1.900 24.0 4.60 0.460
NN
3.0b6		 Pentium
133 MHz		 0.030 0.013 0.031 0.053 0.045 0.045 0.015 0.023 0.053 0.099 14.0 0.120 7.3 0.38 0.076
NN
3.0b7		 Pentium
133 MHz		 0.030 0.016 0.030 0.054 0.046 0.046 0.014 0.023 0.054 0.052 15.0 0.097 8.1 0.25 0.031
IE
3.0b2		 Pentium
133 MHz		 0.030 0.00* 0.015 0.030 0.022 0.022 0.00* 0.022 0.052 0.120 8.1 0.075 2.4 0.34 0.061
NN
2.0		 R4600
132 MHz		   0.390 0.600 0.810 0.820 0.830 0.200 0.790 0.710 0.790 16.0 1.500 12.0 3.10 0.390
NN
2.02		 21064
125 MHz		 1.000 0.520 0.800 1.100 1.700 1.300 0.290 0.220 1.200 1.200 28.0 2.600 22.0 7.30 0.750
NN
2.02		 UltraSPARC
143 MHz		 0.630 0.340 0.540 0.730 0.870 0.900 0.200 0.780 0.690 1.000 5.5 1.100 5.90 3.20 0.360
JDK
1.0.1		 UltraSPARC
167 MHz		 0.500 0.240 0.510 0.560 0.610 0.620 0.140 0.660 0.520 0.560 3.7 0.900 7.5 2.10 1.200
Guava
(proto)		 UltraSPARC
167 MHz		 0.024 0.00* 0.006 0.048 0.048 0.048 0.036 0.066 0.160 0.280 45.0 0.200 4.5 1.00 0.830

Contributors

http://www.cs.cmu.edu/~jch/java/microbenchold.html Java Microbenchmark Results
Last modified: Wed 21 May 1997 Copyright © 1996 Jonathan Hardwick