-
InfluxDB
Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.
Thank you a lot for your thoughts. I have implemented some vectorization into my pet project which I do not want to advertise but I will put some links to some relevant sources. In short I wanted to introduce Vector API into the linear algebra part. I chose the following design: implement a base class to offer default standard implementations. From there I derived three vector classes, one dense, one with stride and another one with indexes. To avoid implementing in each class I made them to implement loadVector and storeVector at some given position. Those are simple methods which basically translated to code like DoubleVector.fromArray(species, array, offset_pos) etc. I tested those implementations against vanilla vector API and I see no issue and there is no memory increase at all. See some vector code here: dense vector, stride vector and some generic code in a base class. I tested with jmh in multiple ways and I see no performance difference. The only hint which I have is that the overloaded methods are inlined by jit since they are small and hot. Perhaps this could be an explanation. What do you think?
Related posts
-
Show HN: Open-source platform for rapid development of enterprise apps
-
Google Blog: Android's theft protection features keep your device and data safe
-
Spring Security with Oauth 2.0
-
Guided Data Access Patterns: A Deal Breaker for Data Platforms
-
Haveno – Decentralized crypto-fiat exchange built on Tor and Monero