GraphReorderAndConverter, source codes in graph-reordering.
see edge_list_to_csr_performance_measure_deg_descending.cpp and edge_list_to_csr_performance_measure.cpp for the OpenMP implementation.
unit: seconds.
| platform | statistics |
|---|---|
| 28-core-cpu | cpu-edge-list-to-csr-performance.md |
| knl | knl-edge-list-to-csr-performance.md |
| exec-name | before csr transform time | before sort time | edge list to csr time |
|---|---|---|---|
| edge_list_to_csr_performance_measure | 1.795s | 5.942s | 9.065s |
| edge_list_to_csr_performance_measure_deg_descending | 1.816s | 5.872s | 11.766s |
hbw for high-bandwidth-memory (MCDRAM), i.e, allocation dst array in the CSR on that multi-channel memory.
| exec-name | before csr transform time | before sort time | edge list to csr time |
|---|---|---|---|
| edge_list_to_csr_performance_measure | 2.822s | 6.953s | 10.494s |
| edge_list_to_csr_performance_measure_hbw | 2.805s | 7.033s | 12.199s |
| edge_list_to_csr_performance_measure_deg_descending | 2.897s | 7.09s | 14.942s |
| edge_list_to_csr_performance_measure_deg_descending_hbw | 2.832s | 7.108s | 16.313s |