What is the new l1_shared_bank_conflict (or the equivalent) counter for CUDA with CC 3.0? In CUPTI User Guide, it says shared_[load|store]_bank_conflict, but it didn’t show when I tried so.
I’m using the command line profiler in Linux.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Counting the number of bank conflicts. | 1 | 1285 | July 19, 2010 | |
| Visual Profiler - l1 shared bank conflicts | 0 | 1292 | April 16, 2012 | |
| Visual Profiler - l1 shared bank conflicts | 2 | 3218 | April 16, 2012 | |
| warp serialization counter in cudaprof 3.0.21 missing? | 3 | 1300 | August 20, 2010 | |
| Shared memory bank conflicts and nsight metric | 15 | 5568 | October 19, 2024 | |
| How to estimate the number of bank conflictions on CC 2.0 | 0 | 2450 | June 4, 2010 | |
| Allocating more share memory than needed resulted in bank conflict | 7 | 149 | October 19, 2024 | |
| detector for bank conflict | 4 | 1417 | May 3, 2012 | |
| Problem about bank conflict test | 6 | 537 | March 12, 2024 | |
| How to profile such metrics like l1tex__data_bank_conflicts_pipe_lsu_mem_global? | 2 | 838 | March 3, 2022 |