This is probably one of the most exciting slides we've had in this class so far (in terms of practical things that can make your eyes pop).
paluri
Reviewing academic articles on super linear speedup is quite fascinating. Super Linear Speedup in DFS is just one of many. It is really cool to see the research on what types of algorithms (it seems like many genetic/biologically based algorithms are!) are prone to the potential of super linear speedup, asside from super linear speedup from practical inhibitions like the working set being greater than size of memory on a single machine, as describe on this slide.
rokislt10
It's mind boggling to think that super-linear scaling can exist. It seems that we are always taught that performance is rarely as good as one would expect it to be, but here is a case where performance is actually better than what we would expect it to be.
uncreative
So, if measure the speedup here with memory constrained scaling we would likely no longer see the super linear speedup. This is because the speedup we are getting here is due to the problem fitting in cache. What might cause us to see super linear speedups when looking at memory constrained scaling?
This is probably one of the most exciting slides we've had in this class so far (in terms of practical things that can make your eyes pop).
Reviewing academic articles on super linear speedup is quite fascinating. Super Linear Speedup in DFS is just one of many. It is really cool to see the research on what types of algorithms (it seems like many genetic/biologically based algorithms are!) are prone to the potential of super linear speedup, asside from super linear speedup from practical inhibitions like the working set being greater than size of memory on a single machine, as describe on this slide.
It's mind boggling to think that super-linear scaling can exist. It seems that we are always taught that performance is rarely as good as one would expect it to be, but here is a case where performance is actually better than what we would expect it to be.
So, if measure the speedup here with memory constrained scaling we would likely no longer see the super linear speedup. This is because the speedup we are getting here is due to the problem fitting in cache. What might cause us to see super linear speedups when looking at memory constrained scaling?