Back to Lecture Thumbnails
kasia4
suninhouse
May not be a direct answer to your question but ``work stealing'' is a commonly technique in parallel computing: https://en.wikipedia.org/wiki/Work_stealing
It appears that there are rigorous math involved and it'd be good to have reference to equations and derivations.
SebL
It looks like thread 1 steals all the rest works and start working on the next one. The following threads will steal the rest afterwards.
blipblop
What does the "cont" variable denote here?
weimin
"cont" means continuation so cont: i=0 means i=1...9 here
stanwie
What is the usage of this STOLEN token?
Please log in to leave a comment.
Copyright 2020 Stanford University
How does thread1 know how much to steal? Why does it steal the rest of the loop and not just part of it? Like if thread 0 is 10% done with a loop and thread 1 steals 90% of it, it seems that isn't the most efficient? Can someone clarify this for me? Thanks!