search

FuTY / mcgpu

Automatically exported from code.google.com/p/mcgpu
1 stars 0 forks source link

Total speed of simulation #25

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
Hello,

I noticed, that the total speed varies widely from one simulation to another. I 
have a Geforce GTX 770 and if I change the phantom or choose different 
source/detector settings my speed varies from about 1 million x-ray/s to 13 
million x-rays/s.
Could you explain this? Could I do any changes (e.g. in the Input-File) to 
optimize the simulation speed? I've tried to change the "GPU-Threads per 
Cuda-Block" and the "simulated histories per GPU-Thread", but I can only make 
it slower.

Best regards
Jenny

Original issue reported on code.google.com by jenny.ko...@t-online.de on 27 Jan 2015 at 5:56

GoogleCodeExporter commented 9 years ago 
Hi Jenny,

The simulation speed is a very trick issue. 

From the GPU point of view the most important point is simulating a lot of 
histories to be able to keep the card busy. If you are running more than 1000 
blocks you should be fine.

If you modify the phantom, the speed might change because your x-rays might be 
absorbed much faster or miss the phantom.

If you change the source-detector distance, you might also get a change of 
speed for simple geometric reasons: lots of x-rays might go through empty space 
outside the phantom, which will appear like super-fast simulation.

In any case, a speed over a million x-rays per second sounds good.

I hope I was able to help!

    Andreu

Original comment by andre...@gmail.com on 28 Jan 2015 at 12:43

GoogleCodeExporter commented 9 years ago 
Thank you, that explains the issue.

Jenny

Original comment by jenny.ko...@t-online.de on 2 Feb 2015 at 2:20