Switch accumulator vectors to atomic<int> vectors and remove locking

cmdevries / LMW-tree

Learning M-Way Tree - Web Scale Clustering - EM-tree, K-tree, k-means, TSVQ, repeated k-means, bitwise clustering

http://lmwtree.devries.ninja

BSD 3-Clause "New" or "Revised" License

75 stars 20 forks source link

Switch accumulator vectors to atomic<int> vectors and remove locking #6

Closed cmdevries closed 9 years ago

cmdevries commented 10 years ago

Also compare performance before and after.

This can just be done by switching the ACCUMULATOR vector type to an atomic version. Note that whole vector operations do NOT need to be atomic, just updating a single dimension in the accumulator vector.

cmdevries commented 9 years ago

This turned out to be 6x slower on Clang / MacOS.