My GPU has 48 KB shared memory per SMX. If two thread blocks are executed simultaneously in a SMX, will the 48 KB divided equally and have 24 KB per thread block? What if I increase the the number of thread blocks?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Usage of shared memory | 11 | 934 | February 1, 2025 | |
| CUDA Device Query says P100 has 48kb shared memory/block... but it's supposed to be 64kb? | 3 | 1265 | June 4, 2017 | |
| shared memory usage per Block VS per SM | 2 | 8687 | May 3, 2010 | |
| Shared memory | 2 | 6971 | April 14, 2011 | |
| Shared memory per block | 1 | 6920 | March 29, 2012 | |
| Question about max shared memory in block and multiprocessor | 1 | 2666 | February 20, 2024 | |
| Lots of blocks, shared memory question | 2 | 823 | April 20, 2015 | |
| shared memory allocation among thread blocks | 3 | 2013 | March 3, 2008 | |
| shared memory size how much shared memory is available | 2 | 661 | January 16, 2011 | |
| So how much shared mem do we really have ? knowing cuda hw better = better optimization | 0 | 1999 | November 20, 2009 |