main > PRAiS

Padding Tuples ? 

> join_init_run

creates task queue ; allocates part of r / s to each thread - split by cache lines

> prj_thread

Pass 1: parallel_radix_partition_optimized

((task = task_queue_get_atomic(join_queue))) => 
  join_function

In essence => K+V - tuple_t - radix join 
K+V K+V 


https://twitter.com/_onurmutlu_/status/1015924702358003712