the metric in question is not necessarily reliable in all cases. There are various other questions on these forums that indicate this, here is one. If you have difficulty assessing bank conflicts using the profiler when you have a case that should not generate bank conflicts, my suggestion is to ask that question on the nsight compute forum.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Bank conflict of tiled matrix multiplication | 1 | 711 | May 17, 2023 | |
| The question of the example of "3.2.2.3 Shared Memory in Matrix Multiplication(C=A*A(T)" i | 0 | 1929 | September 17, 2009 | |
| Matrix Multiplication Bank conflicts problem | 17 | 1989 | August 31, 2018 | |
| Matrix Multiplication and Bank conflicts code included | 3 | 1454 | April 16, 2012 | |
| Shared memory matrix multiplication not working | 6 | 132 | October 11, 2024 | |
| Shared memory bank conflict | 4 | 526 | July 30, 2025 | |
| How to understand the bank conflict of shared_mem | 16 | 14900 | November 19, 2025 | |
| What is "Other Bank Conflicts" | 0 | 59 | February 24, 2025 | |
| Help on analysis & calculate Bank Conflicts of a naive GEMM implementation using CuTe | 0 | 27 | January 15, 2026 | |
| Will this code cause bank conflict ? | 1 | 487 | October 9, 2018 |