Dennis Potter
af25b4b828
This is the version I submitted to RWTH Aachen University at November 9, 2018. |
||
|---|---|---|
| .. | ||
| nodetype_3d_IB | ||
| nodetype_3d_shmem | ||
| nodetype_timer_comparison_w_optimizations | ||
| nodetype_timer_comparison_wo_optimizations | ||
| nodetype_varying_rate_IB | ||
| nodetype_varying_rate_IB_shmem | ||
| nodetype_varying_rate_zeromq_nanomsg | ||
| nodetype_varying_rate_zeromq_nanomsg_shmem_IB | ||
| nodetype_varying_sample_size_IB | ||
| oneway_busy_hist | ||
| oneway_event_hist | ||
| oneway_inline_hist | ||
| oneway_message_size_inline_median | ||
| oneway_message_size_median | ||
| oneway_message_size_wait_median | ||
| oneway_rdma_hist | ||
| oneway_submit_send_comparison_hist | ||
| oneway_unsignaled_inline_hist | ||
| oneway_unsignaled_rdma_hist | ||
| Makefile | ||
| README.md | ||
| plot_histograms.ipynb | ||
| plot_medians.ipynb | ||
| plot_nodetypes.ipynb | ||
README.md
Automatic creation of plots
Cleanup of datasets
In order to save storage and to speed up processing of the datasets, all datasets that are created by the VILLASnode node-type benchmark are cleanup up prior to processing. All actual data is removed, since only the metadata is relevant for the results. Information about the number of values in a sample is contained in the filename.
Samples are removed in the Makefile with:
cut -d, -f3,4 $(file) > temp.csv && mv temp.csv $(file)