Paws.c staging semantics

gistfile1.txt

Stages | Queue ... <added> (Note)
 A B C
[. . .] 
[. . .]<1 2 3>        (Adding three new executions)
[1 2 3]               (No conflicts, so all are staged)
[1 2 3]<4 5>          (Adding two new executions)
[1 * 3] 4 5           (Stage B is finishing, and will acquire a new work unit off the bottom of the queue next)
[1 4 3] 5
[* 4 *]               (A and C finish, but there’s only one piece of work in the queue)
[5 4 .]               (A takes 5, leaving C empty)
[5 4 .]<2 2>          (An old execution is now re-staged, twice! Presumably with new results)
[* 4 2] 2             (C takes one instance of 2, and A clears up)
[. 4 2] 2             (There is nothing stage-able in the queue, so A remains empty, even though the queue is *not* empty)
[. 4 2] 2<6 7 2 5>    (Many new executions added, some new, some old)
[6 * 2] 2 7 2 5       (A takes 6, the bottom-most queued execution that is stage-able, and B finally finishes 4)
[6 7 *] 2 2 5         (B takes 7, and C finishes the current staging of 2)
[* 7 2] 2 5           (C takes a *new staging* of 2, so it continues executing it, and A unstages 6)
[6 7 2] 2 5
[6 7 2] 2 5
[* * 2] 2 5           (A and B unstage)
[5 . 2] 2             (A takes 5, leaving a still-unstagable 2)
[5 . *] 2             (C clears up the previous staging of 2, immediately taking the last staging)
[* . 2]               (The queue is empty, and A finishes, remaining empty)
[. . 2] 
[. . *]               (C finally finishes executing the last re-staging of 2)
[. . .]               (... and with an empty queue and empty stages, the world is complete.)
[. . .] 

Key:
  . → Stage empty
  3 → Stage executing execution-termed-number-three
  * → Stage finished execution