Lecture 24: High Performance Image Processing & Halide (74)
ellenluo
I really enjoyed this slide and how it compares different ways to optimize the same function! I didn't realize that even a simple function like a box blur could have so many different optimizations so I found this especially interesting. I would recommend checking out the webcast to see this in action.
AnastasiaMegabit
I thought the visualization in the webcast was pretty cool too. I would actually like to know if there is an overall "best" or if "it just depends" on the situation.
moridin22
I think the answer is always going to be 'it just depends' since the specs of different systems such as processor speed or cache size will always have a significant effect on which part of the triangle gives you the best tradeoff.
sunsarah
How do you decide between how many of these organizations to test in order to get the best organization vs the overhead to test? Would that also be something user specified?
I really enjoyed this slide and how it compares different ways to optimize the same function! I didn't realize that even a simple function like a box blur could have so many different optimizations so I found this especially interesting. I would recommend checking out the webcast to see this in action.
I thought the visualization in the webcast was pretty cool too. I would actually like to know if there is an overall "best" or if "it just depends" on the situation.
I think the answer is always going to be 'it just depends' since the specs of different systems such as processor speed or cache size will always have a significant effect on which part of the triangle gives you the best tradeoff.
How do you decide between how many of these organizations to test in order to get the best organization vs the overhead to test? Would that also be something user specified?