Comments on: .NET 4.0 And Our Parallel Future

By: Jeff Froom

Jeff Froom — Wed, 13 Jun 2012 02:53:23 +0000

Hello, thanks for the great blog post.

Under what licence have you released this code you wrote that is now listed on this wiki entry?

http://en.wikipedia.org/wiki/Damerau%E2%80%93Levenshtein_distance

MIT/GPL like jquery ? http://jquery.org/license

Thanks!
Jeff

By: Justin Etheredge

Justin Etheredge — Tue, 09 Mar 2010 15:15:41 +0000

@Sam Thanks! And yes, I just ported the code from Wikipedia directly. If I have to use it in production, I’ll most certainly check out your implementation. Actually, I’ll probably check it out for fun anyways. :-)

By: Sam

Sam — Tue, 09 Mar 2010 14:57:13 +0000

Great post, thanks. This was especially interesting for me, as I wrote a parallel processing framework for the purpose of running Levenshtein in .NET 2.0! I will experiment to see if this is any faster, but man, it would have saved me a lot of work back then.

I assume you were using Levenshtein only for the purpose of demo-ing this new framework, but if you actually are trying to optimize, you should use this version of Levenshtein. It’s 2.5 times faster – http://webreflection.blogspot.com/2009/02/levenshtein-algorithm-revisited-25.html

By: Danimal

Danimal — Sat, 27 Feb 2010 23:45:34 +0000

Thanks Justin. It’s pretty cool that with a pretty easy coding change you could get an impressive performance improvement.

By: Justin Etheredge

Justin Etheredge — Sat, 27 Feb 2010 21:19:30 +0000

@Danimal @Rob I have posted the performance numbers that I saw. Unfortunately I only have a dual core proc to try it on.

By: Rob Bazinet

Rob Bazinet — Sat, 27 Feb 2010 12:44:48 +0000

Very interesting Justin. It would have been really cool if you compared the execution times with and without using any parallel extensions and posted them as part of the post.

How significant was the increase in performance?

By: Danimal

Danimal — Fri, 26 Feb 2010 16:36:04 +0000

Great post — thanks. What sort of performance improvements did you see moving to the parallel implementation?