A story of a very large loop with a long instruction dependency chain - Johnny's Software Lab
We at Johnny’s Software Lab LLC are experts in performance. If performance is in any way concern in your software project, feel free to contact us.Some time ago we had to investigate a performance problem for a customer. The problematic piece of code was a loop. The loop was very long, about 1000 lines of vector intrinsics. The vectorization was done efficiently so there were no more efficient intrinsics that we could use to make it faster. Just by looking at it, it seemed that it everything was...
Read more at johnnysswlab.com