Benchmarks

Feel free to add your own Tablix benchmarks in the following table. Please provide a link if you did not use an example XML configuration from the Tablix distribution.

Description of hardware in the cluster (number of machines, operating system, architecture, CPU clock)	Tablix version	XML configuration used	Average GPM
Pentium III, 600 MHz, 256 MB RAM, Debian GNU/Linux 3.1	Tablix 0.3.1	sample2.xml	189
Athlon 1100, 512 MB RAM, OpenSuse GNU/Linux 10.0	Tablix 0.3.1	sample2.xml	405

Description of hardware in the cluster (number of machines, operating system, architecture, CPU clock)	Tablix version	XML configuration used	Average GPM
PowerPC G4 1.33 GHz, Mac OS X 10.3.9	Tablix 0.2.3	sample2.xml	573
single AMD Athlon XP 2500, Debian GNU/Linux, 2.6.11.11 kernel	Tablix 0.2.3	sample2.xml	602
2-way Intel Pentium III 1.00 GHz, Debian GNU/Linux, 2.6.11.12 kernel	Tablix 0.2.3	sample2.xml	684
LASPP cluster, one node	Tablix 0.2.3	sample2.xml	820
2-way Intel Xeon 2.80 GHz (with hyperthreading), Debian GNU/Linux, 2.6.11 kernel	Tablix 0.2.3	sample2.xml	2271
LASPP cluster, three nodes	Tablix 0.2.3	sample2.xml	2410
LASPP cluster, ten nodes	Tablix 0.2.3	sample2.xml	8220

Description of hardware in the cluster (number of machines, operating system, architecture, CPU clock)	Tablix version	XML configuration used	Average GPM
single AMD Duron, 700 MHz, NetBSD/i386 1.6.2	Tablix 0.1.0	sample.xml	345
single AMD Athlon XP 2500, Debian GNU/Linux, 2.6.9 kernel	Tablix 0.1.0	sample.xml	1050
single AMD Athlon XP 2500, Debian GNU/Linux, 2.6.9 kernel	Tablix 0.1.0 with --enable-tunable option	sample.xml	1080
4-way AMD Opteron 2200 MHz, Debian GNU/Linux, 2.6.10-ac2 kernel, pure 64-bit system	Tablix 0.1.0 with --enable-tunable option	sample.xml	6016

Partial results of the LASPP cluster tests are available.

GPM values for Tablix 0.2.3 are comparable to GPM value for Tablix 0.3.0 since no significant changes have been made to the genetic algorithm between these two releases.

Tablix 0.3.1 has fitness caching enabled by default. GPM values are only comparable with earlier versions if fitness caching has been disabled ("cachesize" option set to "0")

---

For those not familiar with Tablix: GPM (generations per minute) number in this table means how many candidate timetables (possible solutions to the timetabling problem) are processed by Tablix in a minute. This number of course depends on the complexity of the timetable that is being generated, parameters of the genetic algorithm (that is why each entry above must state which XML configuration file and which compile options were used) and the computing power of the machine or cluster of machines used. Simple problems usually take around 10000 generations to find a solution.

Because there can be slight differences between genetic algorithms in different versions of Tablix, these numbers are only comparable between the same Tablix release. There is absolutely no way to compare GPM numbers to other timetabling software.