SPEC CPU2006 Benchmark Description

Benchmark Name


Benchmark Author

Cloyce D. Spradling <cloyce+specrand [at] headgear.org>

Benchmark Program General Category

Mine Canary

Benchmark Description

999.specrand is a small harness for the algorithm presented in "Random Number Generators: Good Ones Are Hard To Find" by Park and Miller. The datatypes used have been modified so that the algorithm will work properly on systems with 64-bit longs.

The benchmark simply generates a sequence of pseudorandom numbers starting with a known seed. This benchmark is not a timed component of CPU2006; rather, it's there as an indicator of larger problems. Several of the other benchmarks use the specrand code as their PRNG. Thus, a failure in 999.specrand would point out a source of error in those codes as well. This is cheap (in both time and space) insurance.

Input Description

999.specrand's input consists of two numbers: a seed value for the PRNG, and a count of the numbers to generate.

Output Description

The specified number of random numbers are generated and output twice. The first set is unscaled output from the PRNG output as a standard floating point number with no size or precision modifiers (printf '%f' format). The second set is scaled to between 1 and 2048000, and is output as a standard integer with no size modifiers (printf '%d' format). The PRNG is not re-seeded between sequences, so actually count*2 numbers are generated.

Programming Language


Known portability issues

This code probably will not work on a system where the standard 'int' type is 64-bits wide.


Last updated: 6 February 2006