The leap-frog integrator has been implemented and it yields exactly the same behavior as discussed in the GADGET-2 paper; It’s more reliable than second order Runge-Kutta even though they are both second order methods.  I’ve modified the code so that it initializes particles according to a simple distribution function. The images below are IFrIT plots of  3, 10, and 100 randomly oriented orbits in a simple potential, as well as a plot of 25,000 initial positions.

The main focus of the last few days has been C++ programming and building the individual pieces that will be needed for the N-body code later this month.  I’ve been able to use OpenMP to parallelize the main for-loop.  It’s surprisingly easy.  There is  noticeable speed increase on my dual processor system ( Note: I have not yet benchmarked the program to measure the actual speed-up factor).  The CPU utilization now reaches 100% rather than the previous 50%.  I’m currently working out the bugs with the shared memory system.  Writing a code for parallel execution requires me to modify my way of thinking a bit.  I am also trying to figure out how to handle large amounts of memory in C++.  Once these issues are resolved, I will try to implement a data visualization library so the program can plot output automatically.  This will be important for the next project because I want to be able to make animations of the N-body simulations.

In terms of reading, I just started reading Bruno’s  paper dealing with continuous stellar mass-loss in N-body models.

This entry was posted on Monday, August 10th, 2009 at 9:25 pm and is filed under regular update. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.