Function canonical_warp_idx

include/common.h:496–496

Returns a warp index in the CTA. The threads in warp may not be convergent As it doesn't sync the warp, it faster and allows forward progress

/// Returns a warp index in the CTA. The threads in warp may not be convergent
/// As it doesn't sync the warp, it faster and allows forward progress
DEVICE int canonical_warp_idx() { return threadIdx.x / WARP_SIZE; }