What is Warp divergence?

The splitting of threads within one warp onto different paths at a conditional branch. The paths execute one after another, lowering efficiency.

Field: GPU Programming

Questions that mention Warp divergence (1)

  1. Within one warp, a conditional branch sends half the threads to the if path and half to the else path (warp divergence). How is this generally executed?

Related terms

Bank conflict Constant memory CUDA GPU kernel GPU occupancy GPU shared memory Grid-stride loop Latency hiding Memory coalescing Mixed precision

Back to the glossary