Hi, My name is Anukesh and as a part of my research I am using GNU parallel to do parallel processing with a huge number of data sets. We also have access to HP compute nodes of Purdue University. I tried to execute some program both in HP compute node (20 cores) and a normal desktop computer with 8 threads. However it seems to me that the same program of execution completes much faster in normal desktop (<2 mint) than HPC (around 10 or more mint) with parallel. I also observed while executing GNU parallel in HPC; there it shows impi over taken the program to execute the files. But I do not see such messages in Desktop computer. I am attaching the bash script and output from HPC ( please see this link for files https://drive.google.com/open?id=0B_R9Wj6444c2dzAwc1QtM2hULWM). I execute the qsub with a default of 30 min to check how the program is executing. Could you please sort out what is wrong in this program and I was not sure whether impi interface is delaying this parallel process. It could be very helpful to explain if that the case how to bypass impi and run the code which as same as in desktop computer.
With Regards Anukesh
