I didn't notice the variances before. It seems that the first run is always the slowest and that the other runs are either 2000 ms or 1000 ms but 2000 comes up more frequently than 1000. I have seen this on other benchmarks too where I run several benchmarks and every once in a while I get a real killer of a number.
I suspect that there is a data structure that is used a lot and that it gets aligned on some runs and that this pushes performaance into the stratosphere.
A way to confirm this would be to look at the source code to try to align the suspect structure. I imagine a code profile on the program would be able to find the code given the big difference in CPU time between runs. I have a lot of other stuff to do right now though.
This is using Microsoft's Compiler that generates 64-bit code (not exactly sure which one as I have four (I think) Microsoft compilers on my system right now). I just used the setup command for my Mozilla 64-bit builds and did a cl {modules} -O2.
D:\Download\tscp181\tscp181>main
Tom Kerrigan's Simple Chess Program (TSCP) version 1.81, 2/5/03 Copyright 1997 Tom Kerrigan
"help" displays a list of commands.
tscp> bench
8 . r b . . r k . 7 p . . . . p p p 6 . p . q p . n . 5 . . . n . . N . 4 . . p P . . . . 3 . . P . . . P . 2 P P Q . . P B P 1 R . B . R . K .