[detail] I profiled my customized GEMM kernel, and in the detailed page, Nsight Compute told me that there’s no bank conflict. But in the source view, it told me that the LDGSTS instruction has 50% of excessive shared wavefronts. So I wonder is it possible that excessive wavefronts originate from …

Is it possible that excessive wavefronts originate from sources other than bank conflicts?

ryaaan July 1, 2025, 2:29pm 2

Here’s the source view page.

Topic		Replies	Views
Questions about “L1 Conflicts Shared N-way” & metrics related to “Excessive” Nsight Compute	0	1598	March 29, 2024
Problems about Profiling Shared Memory Bank Conflicts using nsight-compute Nsight Compute	2	1685	January 25, 2022
Questions about "L1 Conflicts Shared N-way" & metrics related to "Excessive" CUDA Programming and Performance	6	681	July 1, 2025
[Nsight Compute] [H100] Questions on L1 Bank Conflict statistic discrepancies between Details and Source pages Nsight Compute kernel , nsight	2	94	December 16, 2025
Analyzing the bank conflicts in my kernel Nsight Compute	2	123	September 30, 2025
[Nsight Compute] [H100] Questions on L1 Bank Conflict statistic discrepancies between Details and Source pages Nsight Compute	3	71	December 26, 2025
Find L1 Wavefronts Shared Excessive in NCU commandline Nsight Compute cuda , nsight	1	897	August 24, 2023
Shared memory bank conflicts and nsight metric CUDA Programming and Performance	15	6110	October 19, 2024
Clarification: bank_conflicts metric vs wavefronts for shared memory LDS.128 CUDA Programming and Performance	1	26	February 6, 2026
Shared memory bank conflict CUDA Programming and Performance	4	526	July 30, 2025