Hi , i am profiling a really simple kernel and there is no shared memory use. __global__ void stride(const float *A, float *B){ int idx = threadIdx.x+blockDim.x*blockIdx.x; B[idx]= A[idx*Stride]; } In memory section of nsight compute, why there are some shared memory use ? [image] l1tex…

Why databank read is non zero, even i do not use any shared memory?

felix_dt February 27, 2025, 12:37pm 2

This was already answered here.

Topic		Replies	Views
The increase of the shared memory size leads to the bankconflict (from 9 KB shared memory) Nsight Compute	5	547	July 14, 2023
Some confusion on using shared memory. CUDA Programming and Performance	26	9136	June 2, 2009
Number of kilobytes transferred to/from shared memory twice the expected CUDA Programming and Performance	12	717	September 29, 2018
Shared memory problem CUDA Programming and Performance	10	3974	April 20, 2010
shared memory example to be found easy example for vector dot product CUDA Programming and Performance	18	3671	September 17, 2010
Using Shared Memory in CUDA C/C++ Technical Blog	36	2043	October 8, 2020
__syncthreads and shared memory CUDA Programming and Performance	21	4411	June 15, 2011
Optimized shared memory index mapping function CUDA Programming and Performance	16	1380	March 20, 2023
How to interpret the difference between LSU utilization and Shared Memory utilization (in case of shared memory access only)? Nsight Compute cuda , kernel	0	579	July 6, 2022
shared memory Computation become slower when using the shared memory CUDA Programming and Performance	8	1837	August 20, 2010