04-13-2015, 03:14 AM
(This post was last modified: 04-13-2015, 03:15 AM by AltruismAndCake.)
I decided to test my idea with actual code, and it turns out that it's not so easy in implementation. For my test I used a 3.8GHz AMD processor for one thread and my Nvidia GPU's 700MHz cuda cores with many threads. In the end I realized that the parallel/threaded code could never compete with a single thread program, because the multithreaded code requires checking if Fibonacci terms already exist, plus the over head of branching.
If anyone has any recommendations, I'll listen, but I think the overhead for trying to run a multi-threaded Fibonacci solver is too high and the simplest way is to simply just add.
Edit fixed typo
If anyone has any recommendations, I'll listen, but I think the overhead for trying to run a multi-threaded Fibonacci solver is too high and the simplest way is to simply just add.
Edit fixed typo