nmslib / hnswlib

Header-only C++/python library for fast approximate nearest neighbors
https://github.com/nmslib/hnswlib
Apache License 2.0
4.34k stars 643 forks source link

It will be very slow when I add 30000 dim 720 features into it #385

Open jonlu5508 opened 2 years ago

jonlu5508 commented 2 years ago

I have init a HierarchicalNSW withm space=L2,maxelements=30000, M=16, ef_construction=200 ef=100 rand_seed=100. and then call addPoint to add feature, at first, it is very fast, but when it goes to 200, it will be slow 7ms,10ms,20ms,40ms,60ms The hardware system is cortex-a17 (arm32) 1.6GHz, RAM 2GB

Mikey2266 commented 2 years ago

I have init a HierarchicalNSW withm space=L2,maxelements=30000, M=16, ef_construction=200 ef=100 rand_seed=100. and then call addPoint to add feature, at first, it is very fast, but when it goes to 200, it will be slow 7ms,10ms,20ms,40ms,60ms The hardware system is cortex-a17 (arm32) 1.6GHz, RAM 2GB

Hi~ Have you solved it yet?

yurymalkov commented 2 years ago

Hi. That is very strange. What is the performance on x86 CPUs with the same data?