Mojo function
elect_one_sync_with_mask
elect_one_sync_with_mask(mask: SIMD[uint32, 1] = __init__[__mlir_type.!pop.int_literal](4294967295)) -> Bool
Elects a single thread within a warp to perform an operation.
Note:
- Only supported on NVIDIA SM90+ GPUs.
- Maps directly to the
elect.sync
instruction in CUDA PTX. - Useful for having a single thread perform an operation while maintaining warp synchronization.
Args:
- mask (
SIMD[uint32, 1]
): The mask to use for the election. Defaults to 0xFFFFFFFF.
Returns:
True for the elected thread, False for all other threads in the warp.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!