If the grid size (the number of thread blocks) in a CUDA kernel is less than the number of streaming multiprocessors (SMs) on the GPU, will some SMs remain idle?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| SM occupancy question | 0 | 554 | September 20, 2011 | |
| How to choose grid size ? No. of blocks and threads ? | 1 | 819 | February 4, 2016 | |
| About Grid size selection | 4 | 1183 | May 17, 2017 | |
| block numbers related to the number of SMs blocks in multiple SMs | 1 | 1411 | December 1, 2009 | |
| Relation between SM and block | 1 | 5592 | March 18, 2010 | |
| Grid Block launch configuration | 2 | 985 | November 18, 2011 | |
| Grid size performance implications | 3 | 2905 | October 10, 2009 | |
| How to choose a proper grid size | 3 | 450 | January 30, 2024 | |
| Grid size limit of concurrent kernels | 6 | 783 | April 5, 2024 | |
| Distribution Threads by the SMs | 1 | 563 | December 15, 2014 |