Hi, all Recently I ran several sets of experiments on MARSSx86 using PARSEC benchmarks, and I found there are some weird behaviors in the performance results.
1. I configured a 4-core machine, then run PARSEC with 1 thread and 4 threads respectively. In the results I found that the scaling of several benchmarks are even over 4 (e.g. canneal), which seems weird to me because I think 4 should be perfect scaling case... 2. I configured another 8-core machine as shown in the MARSSx86 DAC paper and run the same benchmarks (bodytrack, fluidanimate and raytrace) and got different scaling results from the paper's original data... I'm thinking that maybe I did the configuration or benchmark running in a wrong way. I use create_checkpoints.py to create the checkpoints of the benchmarks on the disk image provided by MARSS website and run_bench.py to run the experiments. Does anyone have any thoughts on what could be wrong in the process? Any help will be greatly appreciated!!! If you need more details on the simulation setup, please feel free to email me. Thanks, --Sean
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
