The cost of a board pointer

[Zach Wegner]

I'm converting ZCT to threads, and I've finished everything but the SMP code. Just to lay the argument about there being any sort of speed difference to rest, here is the speed difference between them:

Without board pointer:

Code: Select all

Total nodes=30627886
Total time=35.052
Nodes per second=873784

With board pointer:

Code: Select all

Total nodes=30627886
Total time=35.376
Nodes per second=865781

...So, just under 1 percent, i.e. nothing. Anyone can verify this with my sources if they want.

[bob]

I'm converting ZCT to threads, and I've finished everything but the SMP code. Just to lay the argument about there being any sort of speed difference to rest, here is the speed difference between them:

Without board pointer:

Code: Select all

Total nodes=30627886
Total time=35.052
Nodes per second=873784

With board pointer:

Code: Select all

Total nodes=30627886
Total time=35.376
Nodes per second=865781

...So, just under 1 percent, i.e. nothing. Anyone can verify this with my sources if they want.

I found the same thing when I made the change for SMP years ago. But some will still argue the point even though you actually _did_ it and compared the difference in terms of speed.