If each thread in a block have one float variable each, how do I in an efficient way make a sum of the variables?
kuisma
1
This is called a parallel reduction - take a look at the reduction sample in the SDK.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Accumulate value within block | 15 | 3608 | October 16, 2010 | |
| Adding data from multiple threads | 3 | 3526 | June 20, 2008 | |
| scatter and gather with CUDA? | 3 | 10620 | March 9, 2009 | |
| Question regarding summing up outputs Summing outputs from each thread | 10 | 8254 | March 12, 2008 | |
| accumulating floats accross threads in a block is there and atomicAdd + sync for floats? | 1 | 2322 | January 26, 2009 | |
| Parallel Addition ? How can i serialize parts at kernel? | 4 | 3063 | August 16, 2009 | |
| finding sum | 1 | 2588 | November 18, 2007 | |
| Array Sum in cuda | 5 | 11725 | May 30, 2010 | |
| Simple Inefficient Parallel Addition | 5 | 3331 | April 10, 2009 | |
| Parallel Reduction | 2 | 1285 | July 8, 2010 |