I've written an application which is pretty CPU intensive. It's a simulation, programmed using an object-oriented approach.
I'm trying to optimize the application to run as efficiently as possible. To that end, I've parrallelized as much of the code as possible, and I've parrallelized whatever unavoidable for loops exist in my code.
But even after doing that, my CPU usage, on and 8-core machine, seems "stuck" at ~50%. How do I find what is limiting the application from using more of the CPU?
Here's what I've already done:
-Minimized screen updates, so the CPU doesn't waste time drawing stuff.
-Created as much data-flow path independence as I can,
-Parrallelized for loops.
-VIs in parrellized code are set to shared-close reentrant.
-Inlined as many VIs as I can.
At first I though that the problem was simply that the for loops don't have enough iterations to really take advantage of the parrallellized code, and I can observe that if I scale the number of objects in my simulation to less than about 250 objects (the number of for loop iterations is equal to the number of objects in the simulation). With 250 objects, CPU usage is <25%. If I increase the number of objects to about 1000 objects, CPU usage goes up to nearly 50%. But even after increasing the number of objects even more (say, 10000 objects), CPU usage is still stuck at ~50%.
Any ideas why the application is leaving half the CPU cycles unused? Nothing else is running on the system, and there is no I/O - so no blocking calls. I'm not even streaming data to or from a file. Any help is appreciated.
Thank you!
