I searched the forums for a similar thread but didn't find one, if one already exists please tell me so and move my post.
I did read the dev and saw that you noticed you are working on multithreading this whole thing, I am just adding a few suggestions
Thing is, when I looked into the task manager to see the actual load SE causes, I was kinda suprised that it used ONLY 1 thread(2 when manually setting the loader to the async mode).
Now I have 8 threads for some reason, so I guess you can skip down to OpenCL :v
Now, SE surely is not slow but, wouldn't it boost the generation a lot if you actually created more worker(generator) threads?
I am not too familiar with the method you are using(octrees?) but I am pretty sure that it would be possible to multithread SE into single parts,
sim, graphics, physics and a few generation threads.
Since SE is still in alpha I would assume it would be easier to apply these changes rather quickly.
ANOTHER idea would be taking a look at OpenCL, which seems perfect since SE generates a lot of independant objects and I bet OpenCL could boost SE to a multiple of it's current speed, this however is depedant on the individual GPU the client uses.
Most of the current GPUs that are currently in use support OpenCL and even if it doesn't work you can always fall back.
Added (23.03.2013, 21:26)
Accidently posted in the wrong section, can someone move the thread?