tvm_tirx.backend.cuda.op.warp_min

tvm_tirx.backend.cuda.op.warp_min(value, width=32)

Convenience wrapper: warp_reduce(value, "min", width).