"Can prove that work queue storage for system with T threads is no more than T times that of stack storage for single threaded execution"
Is this because in this approach no thread will spawn tasks and keep on its task stack? It will just keep continuation on its stack. This overhead from memory perspective will be same as the single threaded, where thread runs child first and keep continuation on task stack. So, in worst case, we will have roughly around #threads * per thread memory overhead (which will be for keeping continuation).
Is this the concept which the above mentioned point trying to convey?
pdp
@mak: I think so too.
kapalani
Shouldn't Thread 1 work queue have cont:i =2 since thread 1 is executing foo(1)
"Can prove that work queue storage for system with T threads is no more than T times that of stack storage for single threaded execution"
Is this because in this approach no thread will spawn tasks and keep on its task stack? It will just keep continuation on its stack. This overhead from memory perspective will be same as the single threaded, where thread runs child first and keep continuation on task stack. So, in worst case, we will have roughly around #threads * per thread memory overhead (which will be for keeping continuation).
Is this the concept which the above mentioned point trying to convey?
@mak: I think so too.
Shouldn't Thread 1 work queue have cont:i =2 since thread 1 is executing foo(1)