No, all kernels must be launched from the host. However, kernel launches are asynchronous, so you can queue them up to run in sequence without further host intervention.
seibert
21
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Sharing a single counter (variable) across multiple thread(s) block(s) | 13 | 4688 | December 27, 2017 | |
| Writes in same memory location Cant add numbers from different threads? | 46 | 26232 | July 5, 2007 | |
| Newbie - Need to use shared mem? | 27 | 15429 | December 17, 2008 | |
| Texture Memory / Large Data / Global Memory Advice | 14 | 10984 | May 18, 2010 | |
| __syncthreads and shared memory | 21 | 4765 | June 15, 2011 | |
| [SOLVED] Code his own shared memory with device memory! | 15 | 2759 | October 7, 2015 | |
| Threads, branching and writing to global memory | 3 | 783 | October 24, 2018 | |
| Writing results into global array for only some threads | 5 | 1830 | April 6, 2009 | |
| Wishlist Place your considered suggestions here | 201 | 206997 | April 13, 2009 | |
| cuda shared memory usage + no reduction with threads | 5 | 1206 | April 23, 2012 |