Loop Perforation (2019)
cs.cornell.edu
Loop Perforation (2019)
1–5 of 5 posts
Re: Loop Perforation (2019)
#2 noRe: Loop Perforation (2019)
#3I think this is a very elaborate joke, though I was always surprised by some linear algebra algorithms that had a 'cooking recipe' feeling (looking at you svd) of 'loop enough time to get better results'...
Re: Loop Perforation (2019)
#4When your operations are statistical in nature, as Approximate Computing is, so then these operations themselves are different sampling methods.
If the for loop is calculating a sum, but that value only needs to be over a threshold, there is no reason to add the whole thing, you need a proper sample size and distribution and that it what Loop Perforation is giving you in this case.
Re: Loop Perforation (2019)
#5The way you add the integers up to n is
n(n+1)/2.
Fixed that for you. No loops, and it's exact. Pick a better example.