Increase Linux kernel computation speed

[pepi]

Hi to all
I attend to make special kernel build made only for one purpose: to compute mathematical problems faster.
Please ,can someone verify that is correct optimization I can made in kernel

- Try BFS, for a low end machine it might be better for multi-core tasks. []The kernel patch homepage of Con Kolivas
Or
- Build a custom kernel and set:
Always use Transparent Huge Pages
Disable paravirtualization
100 HZ
No Forced Preemption
SLAB
Pick optimizations for your cpu type

Of course any other idea or link is highly welcome

[fuller]

if you are sincere on that task, you will want to construct an apropriate test case, build a kernel for each of your tweaks and measure and compare the performance.

i predict no more than 5% difference if you dont handle really complex datasets